diff --git a/docs/knowledge-base.md b/docs/knowledge-base.md
index 1dc5adfb3..4e9024bd5 100644
--- a/docs/knowledge-base.md
+++ b/docs/knowledge-base.md
@@ -14,7 +14,7 @@ separate work, not an implemented capability of this knowledge base.
 ```text
 src/inferrsi/data/skill_library/
   README.md                         Library navigation
-  skill-index.json                   Stable IDs → categorized paths/backend (50 skills)
+  skill-index.json                   Stable IDs → categorized paths/backend (51 skills)
   hardware/{ascend,cuda,host}/<id>/   Hardware/platform-specific complete skills
   frameworks/<framework>/<id>/      Framework-owned complete skills
   repositories/<repo>/              Engine entry, sub-skills and focused complete skills
@@ -168,18 +168,18 @@ Complete procedures, sidecars, references and cases are packaged, with 13 Ascend
 repositories, methods or workflows. The 21 engine and research-router skills
 use the same ownership layout under `repositories/<engine>/` and
 `workflows/inference-engine-research/`, including their sub-skills and references.
-The library contains 71 `SKILL.md` files, all in English.
+The library contains 72 `SKILL.md` files, all in English.
 The collection overview, import notes and source manifest belong to the research
 workflow's references; Strata design-space analysis belongs to Strata. The library
 root contains only navigation and the registry.
 
 The [generated catalog](../src/inferrsi/data/repo_packs/inferrsi/knowledge/README.md)
-projects the 50 registered skills into the contribution catalog. Its records
+projects the 51 registered skills into the contribution catalog. Its records
 use `inferrsi.<backend>.skills.<name_with_underscores>` IDs, with `ascend` and
 `cuda` for those hardware-specific scopes and `common` for the remaining skills.
 Each `skill_documents.original` points to the full packaged procedure. The
 `skills` CLI retains the original hyphenated skill IDs and defaults to this
-50-ID registry; engine procedures remain available through library navigation
+51-ID registry; engine procedures remain available through library navigation
 and catalog document pointers.
 
 Use `inferrsi knowledge search --repo-type inferrsi --limit 100` in an ordinary
diff --git a/docs/reference-matrix.md b/docs/reference-matrix.md
index 816835271..87ea57d19 100644
--- a/docs/reference-matrix.md
+++ b/docs/reference-matrix.md
@@ -306,3 +306,13 @@ note, not a third inventory. All NPU status for Drama3 remains `unverified`.
 This framework owner does not implement MLX discovery or claim the separate
 MLX-LM application/cache layer. The generated abstract is a discovery projection,
 not an additional source or validated mechanism.
+
+## MLX-LM application source extraction
+
+| Source / revision | Inspected scope | Extraction and limits |
+|---|---|---|
+| ml-explore/mlx-lm @ `ab7f7048dc341ced2d84b1887eeb17d60b0d1b02` (MIT) | `generate_step`/`stream_generate`, BatchGenerator stage/removal and row filtering, cache factory/KV/BatchKV/LRU prefix paths, selected server model-key/batching/stop branches and native test source | [generation-cache-lifecycle](../src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md), with [pinned readset](../src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/source-map.md): three application mechanisms; CODE and unexecuted TEST-SOURCE; empirical validation not_run, no model/service/backend or matched-performance qualification |
+
+MLX-LM owns request/token/cache policy. This source pin does not qualify its
+compatibility with the separately pinned MLX framework or implement discovery.
+The generated abstract is a projection, not independent corroboration.
diff --git a/docs/searchspace-matrix.md b/docs/searchspace-matrix.md
index fd1531a21..c57bab625 100644
--- a/docs/searchspace-matrix.md
+++ b/docs/searchspace-matrix.md
@@ -319,3 +319,14 @@ rows above remain authoritative; the two-matrix contract is unchanged.
 No new measured or green coverage is credited. Stream dependency guidance does
 not add a Communication/Network capability; no SSD or OS control is introduced.
 `common` indexing categorizes framework guidance and certifies no backend.
+
+## MLX-LM application guidance (source-derived)
+
+| Resource | Stage / named mechanism | Canonical owner | Evidence boundary |
+|---|---|---|---|
+| RAM / Host | Observe / Diagnose / Configure / Verify: request UID/stage/cache-row alignment, prefix identity/copy/trim/eviction and cooperative stop | [generation-cache-lifecycle](../src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md) | Logical cache state and allocated storage differ; native model/row/stop behavior unrun |
+| Accelerator | Observe / Diagnose / Verify: prefill cache evaluation and asynchronous decode submission/read boundary | [generation/batch contract](../src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/generation-batch.md) | Application calls MLX primitives; CPU checks cannot certify Metal/CUDA dispatch, quality or serving performance |
+
+No measured or green coverage is credited. Application batching is not a new
+Network/OS capability; no SSD mechanism is added. `common` indexes guidance and
+does not certify backend compatibility.
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/README.md b/src/inferrsi/data/repo_packs/inferrsi/knowledge/README.md
index da2218469..2d19614d2 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/README.md
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/README.md
@@ -1,6 +1,6 @@
 # InferRSI skill discovery index
 
-Generated abstracts for all 50 registered skills in `data/skill_library/`.
+Generated abstracts for all 51 registered skills in `data/skill_library/`.
 Their complete procedures, references and cases ship in the same package.
 `skill-index.json` maps stable skill IDs to categorized directories. The
 reference and searchspace matrices
@@ -15,7 +15,7 @@ These are knowledge abstracts (`--kind knowledge`), with namespaced package
 IDs distinct from explicitly selected `checkout_ref` skills. Ascend/CUDA
 procedures use their corresponding backend; shared methods use `common`.
 
-Source snapshot: [86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1](https://github.com/InferMatrixLab/InferRSI/tree/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library).
+Source snapshot: [32c744c841ab64ac4e106749603da9b1111b726e](https://github.com/InferMatrixLab/InferRSI/tree/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library).
 
 ```bash
 inferrsi knowledge search 'profiling' --repo-type inferrsi --limit 50
@@ -29,7 +29,7 @@ contain the current input paths and bytes (Git checkout filters are respected).
 
 ```bash
 python tools/build_skill_index.py --source-revision <committed-source-revision>
-python tools/build_skill_index.py --source-revision 86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1 --check
+python tools/build_skill_index.py --source-revision 32c744c841ab64ac4e106749603da9b1111b726e --check
 ```
 
 `--check` detects missing, stale and unexpected generated records without writing.
@@ -65,6 +65,7 @@ References and cases ship with their owning skill, not as duplicate knowledge ab
 | [diagnose-vllm-admission-pressure](common/skills/diagnose-vllm-admission-pressure.json) | common | candidate |
 | [draft-acceptance-issue](common/skills/draft-acceptance-issue.json) | common | unspecified |
 | [evaluate-elastic-serving-placement](common/skills/evaluate-elastic-serving-placement.json) | common | candidate |
+| [generation-cache-lifecycle](common/skills/generation-cache-lifecycle.json) | common | candidate |
 | [host-memory-management](common/skills/host-memory-management.json) | common | candidate |
 | [host-overhead-optimization](common/skills/host-overhead-optimization.json) | common | candidate |
 | [host-overhead-profiling](cuda/skills/host-overhead-profiling.json) | cuda | candidate |
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-aclgraph-capture-replay.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-aclgraph-capture-replay.json
index 445c29377..9610b6c9c 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-aclgraph-capture-replay.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-aclgraph-capture-replay.json
@@ -21,8 +21,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json",
@@ -41,7 +41,7 @@
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md": "718be35dc6d93a97e356ed645df899d43e2c37944f4dece8e167efef810fefea",
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json": "e4c02d4ae85e473947e024c38aed121ca8e5cfbfe1091c5792ce58a349248187",
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md": "6ffbfb68cc7e04923b4fbca2a73a7555b7c3b45a6f40819bb5f3213ef05cf88d",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "ascend-aclgraph-capture-replay",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-device-health.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-device-health.json
index e4ed05581..ad817c3c8 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-device-health.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-device-health.json
@@ -23,8 +23,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/metadata.json",
@@ -41,7 +41,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md": "03829039860ece35519f2e4c339a48b5014ec25e07a2c150a3efb6a3a4e35b4a",
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/metadata.json": "2c23209ef418287f1023f53eba4eda66551f09ab102a61c879df27c25365e757",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "ascend-device-health",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-distributed-parallel.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-distributed-parallel.json
index ea885c397..cd43ca99c 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-distributed-parallel.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-distributed-parallel.json
@@ -24,8 +24,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/metadata.json",
@@ -42,7 +42,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md": "4252cc6f984600393b277c45f8c3ef84f5dd0250c13d5b80a2e1b35d3531c520",
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/metadata.json": "843c0a93bf3e61c107c432e05b23a7466ed07b7c061142fed1e6128ca22734cb",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "ascend-distributed-parallel",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-graph-mode.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-graph-mode.json
index bf24dadec..ca7dc9fd1 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-graph-mode.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-graph-mode.json
@@ -22,8 +22,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json",
@@ -40,7 +40,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md": "9203cc5a14a1d65df5fbf275734b93be023e06527459b2f79864f28a338cb44a",
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json": "b0bfa30de94dcda06e2289f5d4a2e44a473f40fb871059e4580f33588f2a51dc",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "ascend-graph-mode",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-model-adaptation.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-model-adaptation.json
index 21619bb51..9e124a864 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-model-adaptation.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-model-adaptation.json
@@ -22,8 +22,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json",
@@ -40,7 +40,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md": "d9e81a63981819fac8f5a21dca9c90f86e26788f58339d3a9c8e66319c178472",
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json": "4c6d0c901f0b3858c7dc337c86d892194e4102c8b349932a4caf1a811c99b1ed",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "ascend-model-adaptation",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-profiling.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-profiling.json
index 7cb5a9279..ea2f569f2 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-profiling.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-profiling.json
@@ -23,8 +23,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json",
@@ -41,7 +41,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12",
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json": "e45210a4df7b1ee183f06573dbc36db5ede182f11ab1cfad5ae22f523b89c557",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "ascend-operator-profiling",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-tuning.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-tuning.json
index b5b988de8..db1217f96 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-tuning.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-tuning.json
@@ -24,8 +24,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json",
@@ -42,7 +42,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592",
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json": "ab7dedfa873b8572db277e43536e6ce04c68ac987aee189ffd5861e4e9bb2d6d",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "ascend-operator-tuning",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-optimization-casebook.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-optimization-casebook.json
index 04c90fe8d..19fc8ee9a 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-optimization-casebook.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-optimization-casebook.json
@@ -22,8 +22,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/metadata.json",
@@ -40,7 +40,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md": "647f5e9a288f2c4e1617ba53df59d9d827df749eb4fe86182d21a9c41175caf8",
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/metadata.json": "b6c4b9a73862b4434038f8e1bc24073864ee334014598fb998af592acb54bcb4",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "ascend-optimization-casebook",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-performance-analysis.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-performance-analysis.json
index a3d02daf0..1bb513f9b 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-performance-analysis.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-performance-analysis.json
@@ -22,8 +22,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json",
@@ -40,7 +40,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md": "955a7b4206e4ea3a73e9aa8723df1ad5f823f950520441bcfdf68d8ba07ca6c7",
     "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json": "ce1ab797865799a58a1bb0ce5ee949a061283b6d5f0353c29ab6c7a95a3a12f1",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "ascend-performance-analysis",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascendc-tiling-design.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascendc-tiling-design.json
index e16db9913..ebc7704c3 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascendc-tiling-design.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascendc-tiling-design.json
@@ -24,8 +24,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json",
@@ -42,7 +42,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b",
     "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json": "04dd3e8961d5c67105b11ff52f1354f34124d9af1b03f9da08cba67cb9f9d463",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "ascendc-tiling-design",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/cann-attribution-overlap.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/cann-attribution-overlap.json
index ac84bdfa4..f13f5fab9 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/cann-attribution-overlap.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/cann-attribution-overlap.json
@@ -24,8 +24,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json",
@@ -42,7 +42,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2",
     "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json": "64b83aae3b9be5ada75e6111114a93bf55173d9df913b4ea8bafd2b4cea23018",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "cann-attribution-overlap",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-device-memory-lifecycle.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-device-memory-lifecycle.json
index 920189376..f5e96500e 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-device-memory-lifecycle.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-device-memory-lifecycle.json
@@ -22,8 +22,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json",
@@ -42,7 +42,7 @@
     "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230",
     "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json": "7ee44d25214f4a6bbeeea4bec61941b2ec2ea469e595865e9346a7589576bd16",
     "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md": "80cba4d3a884e7ee2618326cf74b9f9c3a34e4d9cf443707f6bbe673f2d5ba65",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "torchnpu-device-memory-lifecycle",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-pinned-host-memory-lifecycle.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-pinned-host-memory-lifecycle.json
index b59ebfcec..6805cdde0 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-pinned-host-memory-lifecycle.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-pinned-host-memory-lifecycle.json
@@ -22,8 +22,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/metadata.json",
@@ -42,7 +42,7 @@
     "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/SKILL.md": "23e93ea7c98f99743509fdc04139d78ed2e7123c63e80b366c6beaba7eb70d0d",
     "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/metadata.json": "17f96ff8d36a844d4a62a06975ad918bb8554f343543a3e7ea23e2d98c37fd0b",
     "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/references/source-notes.md": "435193590b4a2ba32e2e4c6fa9d7e8f889528d412cbbb48e35ec5433f7dab384",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "torchnpu-pinned-host-memory-lifecycle",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-serving-preemption.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-serving-preemption.json
index 0cb532b81..444fb9b96 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-serving-preemption.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-serving-preemption.json
@@ -16,8 +16,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md",
         "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json",
@@ -36,7 +36,7 @@
     "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d",
     "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json": "4c52db9b143204496d9bc194d981322ad39a6f8cb399fb49abcb80d7b2924533",
     "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md": "f29fcecfba0490008a354c11b42403983fe165e56ff907ad61cc6040d2b6407f",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "assess-serving-preemption",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-split-backward-schedule.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-split-backward-schedule.json
index 94d3c0f71..5be433a3a 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-split-backward-schedule.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-split-backward-schedule.json
@@ -16,8 +16,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md",
         "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/metadata.json",
@@ -36,7 +36,7 @@
     "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3",
     "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/metadata.json": "858fdd88c6bd2f118bd1d527bb37d47bcc2722cc6140d52ee4657e646cdce080",
     "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md": "61b1609f71c37b61d136b8c4d39f793de87cffb1cf31c8ec737a1c3fa6fb1b4e",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "assess-split-backward-schedule",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-vllm-kv-lifetimes.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-vllm-kv-lifetimes.json
index 8cfaa31e4..30cf5d932 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-vllm-kv-lifetimes.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-vllm-kv-lifetimes.json
@@ -21,8 +21,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/SKILL.md",
         "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/metadata.json",
@@ -41,7 +41,7 @@
     "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/SKILL.md": "313733a65cc48442b18dda1c4665139a689963a90e1032a2235bc6f40e3a8e06",
     "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/metadata.json": "d061abfaf80db3d1b5b34644e38281f9d623d1f49d1e17ef872ebd1a705d4b6c",
     "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/references/source-notes.md": "3665b034594fc8cc767c408002f4e16f8803369242405027c7f37ce6c9ecebe9",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "audit-vllm-kv-lifetimes",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/calibrate-serving-simulation.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/calibrate-serving-simulation.json
index 67daa7d96..760501d26 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/calibrate-serving-simulation.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/calibrate-serving-simulation.json
@@ -16,8 +16,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md",
         "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/metadata.json",
@@ -36,7 +36,7 @@
     "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352",
     "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/metadata.json": "553406b95177a38d73ea6f305c30f361765a437dfe8cfec86c26d4c00e1ac41c",
     "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md": "9696d3c0ef96b6611787446bd6f96d4d6255ca8d107105de8c1d9af83edb9fa5",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "calibrate-serving-simulation",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cgroup-resource-controls.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cgroup-resource-controls.json
index 1dd8773e9..47b6faa80 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cgroup-resource-controls.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cgroup-resource-controls.json
@@ -25,8 +25,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/metadata.json",
@@ -43,7 +43,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5",
     "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/metadata.json": "e0cf9b8e1b92899da1564df2f2fd5c842fa2113925b188358c4984f5e31712ce",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "cgroup-resource-controls",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/checkpoint-recovery-consistency.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/checkpoint-recovery-consistency.json
index 9234c91ea..39b6e178d 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/checkpoint-recovery-consistency.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/checkpoint-recovery-consistency.json
@@ -22,8 +22,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md",
         "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/metadata.json",
@@ -42,7 +42,7 @@
     "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md": "30d55d7d0475d95f369f7c9aa3e594406a9407ded2a4e623df0c39cd487ba724",
     "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/metadata.json": "5c18a225cd3b88051c57cb29d3d9cca3dcf18826aff86cef1cec39f315a1fc1d",
     "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md": "7c95f3058bbfeda1dd7e83d05917a94be0548350c95d7066dd5bf49cb232775a",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "checkpoint-recovery-consistency",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/choose-parallel-training-plan.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/choose-parallel-training-plan.json
index 625e1e36f..ebfb7a064 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/choose-parallel-training-plan.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/choose-parallel-training-plan.json
@@ -16,8 +16,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md",
         "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/metadata.json",
@@ -36,7 +36,7 @@
     "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md": "25d04aed9782b2a8ff1186828becdc7fce93c4418039ceaef9b8dcc5ec9bcd24",
     "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/metadata.json": "858fdd88c6bd2f118bd1d527bb37d47bcc2722cc6140d52ee4657e646cdce080",
     "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/references/source-notes.md": "9f9184d46b19312138a6e805e439e43dc0594a0cc6b46675255a582dc946ca5b",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "choose-parallel-training-plan",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cite-check-downstream-pr.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cite-check-downstream-pr.json
index b5fc18889..00b9150bf 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cite-check-downstream-pr.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cite-check-downstream-pr.json
@@ -16,8 +16,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/workflows/cite-check-downstream-pr/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/workflows/cite-check-downstream-pr/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/skill-index.json",
         "src/inferrsi/data/skill_library/workflows/cite-check-downstream-pr/SKILL.md"
@@ -31,7 +31,7 @@
   },
   "checkout_ref": "checkout:cite-check-downstream-pr",
   "source_files": {
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939",
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee",
     "src/inferrsi/data/skill_library/workflows/cite-check-downstream-pr/SKILL.md": "485932cdf3b1fba5dc69be73e75d38f86a7b79442d26e82a296c963a7598d226"
   },
   "skill_metadata": {
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/claim-contract.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/claim-contract.json
index 272db1cc9..312d313cc 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/claim-contract.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/claim-contract.json
@@ -16,8 +16,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/skill-index.json",
         "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md"
@@ -31,7 +31,7 @@
   },
   "checkout_ref": "checkout:claim-contract",
   "source_files": {
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939",
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee",
     "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55"
   },
   "skill_metadata": {
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/comfy-ui.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/comfy-ui.json
index 742c97df5..3658c7d5b 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/comfy-ui.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/comfy-ui.json
@@ -22,8 +22,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/repositories/comfy-ui/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/repositories/comfy-ui/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/repositories/comfy-ui/LICENSE",
         "src/inferrsi/data/skill_library/repositories/comfy-ui/SKILL.md",
@@ -50,7 +50,7 @@
     "src/inferrsi/data/skill_library/repositories/comfy-ui/references/model-residency.md": "014c68386e7cdfdb7b4f566836e8b2ff61bd5a5a506e7b4723f83fe80c5d4de7",
     "src/inferrsi/data/skill_library/repositories/comfy-ui/references/repo-provenance.md": "d7ca20c50652ef70da198de9216315b98cfbaa28b22bfe036f293a85a8f20261",
     "src/inferrsi/data/skill_library/repositories/comfy-ui/references/validation-boundaries.md": "1913dedbeb558d872d6a0e015cc11d822910c3cb259a236db29fb60ad8a64263",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "comfy-ui",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/competitor-scan.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/competitor-scan.json
index 9af4a8889..d6a6ce992 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/competitor-scan.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/competitor-scan.json
@@ -16,8 +16,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/workflows/competitor-scan/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/workflows/competitor-scan/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/skill-index.json",
         "src/inferrsi/data/skill_library/workflows/competitor-scan/SKILL.md"
@@ -31,7 +31,7 @@
   },
   "checkout_ref": "checkout:competitor-scan",
   "source_files": {
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939",
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee",
     "src/inferrsi/data/skill_library/workflows/competitor-scan/SKILL.md": "0ede58ace65a9324a538b4816add1684f230ba536d4364fe5a5dd2c0bbed3776"
   },
   "skill_metadata": {
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dcp-async-save-performance.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dcp-async-save-performance.json
index ed4771859..ecb46ea54 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dcp-async-save-performance.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dcp-async-save-performance.json
@@ -23,8 +23,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md",
         "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json",
@@ -43,7 +43,7 @@
     "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845",
     "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json": "2452e400592bc6e084981b27153f6936b5a12e7f3584951cb8e3373e9c27db4e",
     "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "dcp-async-save-performance",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dense-performance-evidence.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dense-performance-evidence.json
index 81a2b3568..268dd4faa 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dense-performance-evidence.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dense-performance-evidence.json
@@ -21,8 +21,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md",
         "src/inferrsi/data/skill_library/methods/dense-performance-evidence/metadata.json",
@@ -39,7 +39,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933",
     "src/inferrsi/data/skill_library/methods/dense-performance-evidence/metadata.json": "d8cfc7c93d0a96d5240f085732f3f6a45159c125238a693220687489d6fd1979",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "dense-performance-evidence",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diagnose-vllm-admission-pressure.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diagnose-vllm-admission-pressure.json
index d304e341a..6e7e281d7 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diagnose-vllm-admission-pressure.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diagnose-vllm-admission-pressure.json
@@ -21,8 +21,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/SKILL.md",
         "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/metadata.json",
@@ -67,7 +67,7 @@
     "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/loongserve-readset.md": "f1a383e7bdbbd5a5a472d7f6e1292c8a17d4910ff736ca5a6aa422e1bee74f8c",
     "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/orca-readset.md": "a497ef23b7e60c0fbc15d95b7e07cd20547cfe56f37c24dfe6ed70d79ef7a3cf",
     "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/source-notes.md": "84033fd514773f10c0bddd9914fabb980497ba2b308f226a922c3e1888413827",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "diagnose-vllm-admission-pressure",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/draft-acceptance-issue.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/draft-acceptance-issue.json
index 65ed4fa71..b3238a683 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/draft-acceptance-issue.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/draft-acceptance-issue.json
@@ -16,8 +16,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/skill-index.json",
         "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md"
@@ -31,7 +31,7 @@
   },
   "checkout_ref": "checkout:draft-acceptance-issue",
   "source_files": {
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939",
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee",
     "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56"
   },
   "skill_metadata": {
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/evaluate-elastic-serving-placement.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/evaluate-elastic-serving-placement.json
index a0bf442ca..7a732b3b5 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/evaluate-elastic-serving-placement.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/evaluate-elastic-serving-placement.json
@@ -16,8 +16,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md",
         "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json",
@@ -36,7 +36,7 @@
     "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc",
     "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json": "9efdc98c82784f818994d9edc3ce5411ea38c7904f81ceefa15357dc1c94bd2e",
     "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md": "ea416831fd7d0874e06d0a76fb0c7196df53072b717673a0cdf635db6bfd917b",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "evaluate-elastic-serving-placement",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/generation-cache-lifecycle.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/generation-cache-lifecycle.json
new file mode 100644
index 000000000..b411b88e9
--- /dev/null
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/generation-cache-lifecycle.json
@@ -0,0 +1,119 @@
+{
+  "schema_version": 1,
+  "id": "inferrsi.common.skills.generation_cache_lifecycle",
+  "title": "generation-cache-lifecycle",
+  "repo": "inferrsi",
+  "tier": "contribution",
+  "backend": "common",
+  "component": "skills",
+  "scenario": "skills.discovery",
+  "status": "imported",
+  "summary": "Trace MLX-LM prefill, generation batching, prompt-cache reuse and cooperative stopping when request state, cache rows or retained prefixes become inconsistent.",
+  "tags": [
+    "mlx-lm",
+    "generation",
+    "batching",
+    "kv-cache",
+    "cancellation"
+  ],
+  "origin": [
+    "external"
+  ],
+  "sources": [
+    {
+      "type": "git",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
+      "paths": [
+        "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md",
+        "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/metadata.json",
+        "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/cache-stop.md",
+        "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/generation-batch.md",
+        "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/source-map.md",
+        "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/validation.md",
+        "src/inferrsi/data/skill_library/skill-index.json"
+      ],
+      "content_sha256": "bf1177cb9985c12a9bf11edb64771018e1e314c4713920de8fd4a989ff70115f"
+    }
+  ],
+  "index_type": "packaged-skill",
+  "skill_documents": {
+    "original": "data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md"
+  },
+  "checkout_ref": "checkout:generation-cache-lifecycle",
+  "source_files": {
+    "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md": "bf1177cb9985c12a9bf11edb64771018e1e314c4713920de8fd4a989ff70115f",
+    "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/metadata.json": "2ce3abb080bf8494bb99178576ca7d73f10f858b90153766a288d1dd7b599574",
+    "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/cache-stop.md": "bba6b5c73261065309cbcf11b1e591eef48916f0c79ceaaf74d98a9fa96b0992",
+    "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/generation-batch.md": "f16d3c57d45e63bcaffd3871aa41c76972ee11f7dbf1e12b45d9b170303cad9b",
+    "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/source-map.md": "e3d815d4307b5eb3a61545ecee6e56208349755fb9a0ee530bb112b90af60e7f",
+    "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/validation.md": "73b5dd347017baba519ecc85415aead9993c1751d116b9ba4f07689f58c5f783",
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
+  },
+  "skill_metadata": {
+    "name": "generation-cache-lifecycle",
+    "description": "Trace MLX-LM prefill, generation batching, prompt-cache reuse and cooperative stopping when request state, cache rows or retained prefixes become inconsistent.",
+    "profile": "local-skill-v1",
+    "schema_version": 1,
+    "kind": "skill",
+    "origin": [
+      "external"
+    ],
+    "status": "candidate",
+    "empirical_validation": "not_run",
+    "license": "MIT",
+    "tags": [
+      "mlx-lm",
+      "generation",
+      "batching",
+      "kv-cache",
+      "cancellation"
+    ],
+    "sources": [
+      {
+        "type": "git",
+        "url": "https://github.com/ml-explore/mlx-lm",
+        "commit": "ab7f7048dc341ced2d84b1887eeb17d60b0d1b02",
+        "paths": [
+          "mlx_lm/generate.py",
+          "mlx_lm/models/cache.py",
+          "mlx_lm/server.py",
+          "tests/test_generate.py",
+          "tests/test_prompt_cache.py"
+        ],
+        "readset": "references/source-map.md"
+      }
+    ],
+    "applicability": {
+      "scope": "MLX-LM generation, batch-row and prefix-cache diagnosis; model/backend/runtime compatibility requires separate qualification."
+    },
+    "limitations": [
+      "Source-derived candidate; no local native-backend, model, service or efficacy validation.",
+      "Static content checks do not promote this candidate or establish installed API conformance."
+    ],
+    "provenance_readsets": [
+      "references/source-map.md"
+    ],
+    "references": [],
+    "consumption_mode": "packaged"
+  },
+  "applicability": {
+    "scope": "MLX-LM generation, batch-row and prefix-cache diagnosis; model/backend/runtime compatibility requires separate qualification."
+  },
+  "recommendations": [
+    "Read the pinned SKILL.md and its prerequisites before choosing this procedure.",
+    "Read the packaged references and owner-local cases alongside the procedure."
+  ],
+  "limitations": [
+    "This generated abstract points to the complete packaged skill; external evidence and upstream repositories remain source links.",
+    "The imported status describes this index, not the original skill; its declared status is candidate.",
+    "Reading or indexing does not authorize execution, validate efficacy, promote a candidate or grant a license.",
+    "The backend categorizes the procedure's hardware dependency; it does not certify compatibility, and skills.discovery is a discovery scenario.",
+    "Source-derived candidate; no local native-backend, model, service or efficacy validation.",
+    "Static content checks do not promote this candidate or establish installed API conformance."
+  ],
+  "verification": {
+    "locally_verified": false,
+    "evidence_level": "guidance"
+  }
+}
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-memory-management.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-memory-management.json
index ef4822bd8..e05fa5f5b 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-memory-management.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-memory-management.json
@@ -25,8 +25,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/host/host-memory-management/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/host/host-memory-management/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/host/host-memory-management/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/host/host-memory-management/metadata.json",
@@ -43,7 +43,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/hardware/host/host-memory-management/SKILL.md": "b42e565be94d60e0a1557d5ed76a0548e89b8e054dc249c46215e7b90fe4277f",
     "src/inferrsi/data/skill_library/hardware/host/host-memory-management/metadata.json": "f98060ed3c2ef815832eae549bfb21ca02811ff980a629bf995a5b8e7f48c7cd",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "host-memory-management",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-overhead-optimization.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-overhead-optimization.json
index 36f29b827..bc1674021 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-overhead-optimization.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-overhead-optimization.json
@@ -22,8 +22,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/metadata.json",
@@ -40,7 +40,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md": "db8357dab79cfa474561732849b5bb153222f63e6688b6f873f45800595f70b8",
     "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/metadata.json": "90941490debf1ec9f02310fa033658ec294a8e968783d4bb9d004bcded0f73fe",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "host-overhead-optimization",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-benchmark-design.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-benchmark-design.json
index 70cf2a27f..d80d42cf3 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-benchmark-design.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-benchmark-design.json
@@ -21,8 +21,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/methods/inference-benchmark-design/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/methods/inference-benchmark-design/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/methods/inference-benchmark-design/SKILL.md",
         "src/inferrsi/data/skill_library/methods/inference-benchmark-design/metadata.json",
@@ -39,7 +39,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/methods/inference-benchmark-design/SKILL.md": "dcaa6d7bdcbe4c29116b9679d95ec076faaada3e66c647e95acdd46c2343a2eb",
     "src/inferrsi/data/skill_library/methods/inference-benchmark-design/metadata.json": "1e04bcca099b0cdbcbdfd8e9c88b47c1accb656cb5317ccb7a929fcacb0a0d87",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "inference-benchmark-design",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-optimization-casebook.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-optimization-casebook.json
index e20698bf4..e7d8902eb 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-optimization-casebook.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-optimization-casebook.json
@@ -21,8 +21,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/methods/inference-optimization-casebook/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/methods/inference-optimization-casebook/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/SKILL.md",
         "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.json",
@@ -55,7 +55,7 @@
     "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.json": "a2df4056af372fbbb2fbe90db01acc926fad78775844fa91cff8bc2f3286ff58",
     "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.md": "fdcadb4488e29ca45edf21a350ed28ed4f6c6a9fb4772a61d90ca8f62ce5bb04",
     "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/metadata.json": "6aa0dd8ae9a8b45d770593f236ce39a6c2e06a8e67ad7176b47d484686028f63",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "inference-optimization-casebook",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kernel-writing-guide.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kernel-writing-guide.json
index 482285459..c45d475c1 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kernel-writing-guide.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kernel-writing-guide.json
@@ -24,8 +24,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md",
         "src/inferrsi/data/skill_library/methods/kernel-writing-guide/metadata.json",
@@ -42,7 +42,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md": "e238a3102b06af1086bb805f1c3f0f24aacaa03032760e0d95231bcf25878dd1",
     "src/inferrsi/data/skill_library/methods/kernel-writing-guide/metadata.json": "7e82c7b5bcbd695981eaf0ce3ea45fe37dd78c43a73204e1465af5379a205bfb",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "kernel-writing-guide",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kv-tier-management.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kv-tier-management.json
index 5aeebc5e6..db9bca989 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kv-tier-management.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kv-tier-management.json
@@ -24,8 +24,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md",
         "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json",
@@ -42,7 +42,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27",
     "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json": "20f7989a27a438170e4f59eb6bf0691dec25bb3ea7bcb8cc3d364e48b9a698bd",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "kv-tier-management",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lazy-evaluation-stream-lifetimes.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lazy-evaluation-stream-lifetimes.json
index 6d67ea316..75d695496 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lazy-evaluation-stream-lifetimes.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lazy-evaluation-stream-lifetimes.json
@@ -22,8 +22,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md",
         "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json",
@@ -46,7 +46,7 @@
     "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md": "c4bc41e08aca68f661325f61e6f0db1760a079dc362f295c847bdc374fee3bf6",
     "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md": "ccbc70ece4e910b5f3734617cd189fb61b6bb75bd652d2add7a59663d478f83e",
     "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md": "ebfc00bf7d6ab0a1dc415953ae84d56ccdc713d5614aff85e433c4a843793bff",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "lazy-evaluation-stream-lifetimes",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/npu-skill-factory.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/npu-skill-factory.json
index cd2c2e2c2..29c88e16e 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/npu-skill-factory.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/npu-skill-factory.json
@@ -22,8 +22,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/workflows/npu-skill-factory/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/workflows/npu-skill-factory/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/skill-index.json",
         "src/inferrsi/data/skill_library/workflows/npu-skill-factory/SKILL.md",
@@ -54,7 +54,7 @@
   },
   "checkout_ref": "checkout:npu-skill-factory",
   "source_files": {
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939",
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee",
     "src/inferrsi/data/skill_library/workflows/npu-skill-factory/SKILL.md": "74cd863df2e80dc2bda1713d24207809173a3116f30325e3487e1e3517cfec1e",
     "src/inferrsi/data/skill_library/workflows/npu-skill-factory/metadata.json": "8d82e5e7bf2eece16056076e1ab888669330c2533fee26fabe747fbb551e1388",
     "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/COMPARISON.md": "38587678a96581f5c2dc697d982127343565504bffa191352811dffcc165aa23",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/optimization-search-policy.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/optimization-search-policy.json
index b4694426b..1b0b1643a 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/optimization-search-policy.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/optimization-search-policy.json
@@ -21,8 +21,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/methods/optimization-search-policy/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/methods/optimization-search-policy/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/methods/optimization-search-policy/SKILL.md",
         "src/inferrsi/data/skill_library/methods/optimization-search-policy/metadata.json",
@@ -51,7 +51,7 @@
     "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/knowledge/adaptation-interfaces.md": "5e761957cb5ecd93926e6ffc13deee7c830c6e8f0ee3c3846e9dd3b50dc74446",
     "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/references/source-notes.md": "54d83c3713045d59a48821bcf93598374ba4a630e3891ad39498f919317c4fb8",
     "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/updates/optimization-search-policy.md": "d3db44fa9a6831b86a02c638b323ed943e69fe256866213f10d68c72d762879e",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "optimization-search-policy",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/os-process-placement.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/os-process-placement.json
index c3787585d..50232e52b 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/os-process-placement.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/os-process-placement.json
@@ -24,8 +24,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/host/os-process-placement/metadata.json",
@@ -42,7 +42,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md": "8d38af7ee1cd08a99bc6981a4ed1891be6331a6a56ea2537364fc0a1b79702b9",
     "src/inferrsi/data/skill_library/hardware/host/os-process-placement/metadata.json": "1d1c3a34333185159ac01280dd86ea930d48c263be5d36c4fa9f026bb7c56c62",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "os-process-placement",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/performance-ceiling-analysis.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/performance-ceiling-analysis.json
index 8231b8015..43eef2d8b 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/performance-ceiling-analysis.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/performance-ceiling-analysis.json
@@ -22,8 +22,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/SKILL.md",
         "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/metadata.json",
@@ -72,7 +72,7 @@
     "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/inferencebench.md": "21f47111197a0b7f8d6384ea918d139dc8318a62ca3313ce060048271f50b639",
     "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/prior-coverage.md": "35d47ac078591221fcb666219bcd824f9400fc17aafecdc0499d8336f7178238",
     "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/vidur.md": "d38602af5219f5feeb49e6cd6bf53d66098c77fad141dae676f1d8a6743cf25d",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "performance-ceiling-analysis",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/plan-activation-residency.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/plan-activation-residency.json
index badc73cdd..0fa78c926 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/plan-activation-residency.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/plan-activation-residency.json
@@ -16,8 +16,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md",
         "src/inferrsi/data/skill_library/methods/plan-activation-residency/metadata.json",
@@ -36,7 +36,7 @@
     "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e",
     "src/inferrsi/data/skill_library/methods/plan-activation-residency/metadata.json": "858fdd88c6bd2f118bd1d527bb37d47bcc2722cc6140d52ee4657e646cdce080",
     "src/inferrsi/data/skill_library/methods/plan-activation-residency/references/source-notes.md": "5ff52655cfd06fcf970ed2a7d4d4fe1c31115b102af2d41497de0c74848ed2b3",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "plan-activation-residency",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/repository-evolution-analysis.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/repository-evolution-analysis.json
index 9489ffc2a..913a155ec 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/repository-evolution-analysis.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/repository-evolution-analysis.json
@@ -21,8 +21,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/methods/repository-evolution-analysis/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/methods/repository-evolution-analysis/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/SKILL.md",
         "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/metadata.json",
@@ -39,7 +39,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/SKILL.md": "8cac8e0d52fa5bffdf3076425c45a6041a3fe4402ffd5219abdf23ccaf338f38",
     "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/metadata.json": "bbde3cec2545a2613653ec521053f972235742fb53b630badcffd9a61fa5f19f",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "repository-evolution-analysis",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/rsi-evidence.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/rsi-evidence.json
index e5ed7a761..3b975c11d 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/rsi-evidence.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/rsi-evidence.json
@@ -16,8 +16,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/workflows/rsi-evidence/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/workflows/rsi-evidence/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/skill-index.json",
         "src/inferrsi/data/skill_library/workflows/rsi-evidence/SKILL.md"
@@ -31,7 +31,7 @@
   },
   "checkout_ref": "checkout:rsi-evidence",
   "source_files": {
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939",
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee",
     "src/inferrsi/data/skill_library/workflows/rsi-evidence/SKILL.md": "a55610874b658b86a39a1a049efb4e4eab0140f6579ed74dd7718a4ca8cedaca"
   },
   "skill_metadata": {
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/skill-evolution-playbook.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/skill-evolution-playbook.json
index a2eefb395..2d76b990c 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/skill-evolution-playbook.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/skill-evolution-playbook.json
@@ -21,8 +21,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md",
         "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/metadata.json",
@@ -39,7 +39,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md": "9cd37f21fe0d862069e4798edca8a04d59f8b840e55d521901594a63dddd9ba3",
     "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/metadata.json": "3132b700334856ab8e1eec4396bcb57627e987adf28cb86d5dc7974820ee9968",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "skill-evolution-playbook",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/storage-io-measurement.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/storage-io-measurement.json
index f31a692ab..51d17e244 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/storage-io-measurement.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/storage-io-measurement.json
@@ -24,8 +24,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json",
@@ -42,7 +42,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7",
     "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json": "1d3195ca8ac8e1b0e94acda01863b83046cff7d773e8131e9f3f925150bad66e",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "storage-io-measurement",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-input-pipeline.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-input-pipeline.json
index bf9575b9d..e2d7406c8 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-input-pipeline.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-input-pipeline.json
@@ -23,8 +23,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md",
         "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json",
@@ -43,7 +43,7 @@
     "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e",
     "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json": "f971cbd0b9ee12afa896227fbc2625c97d307a250a31634c76611da1977855e8",
     "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md": "60e0f3e88655d8e1d5eb4797f39ca4370deda01614f265b8d2ea4559d8ac020f",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "training-input-pipeline",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-resource-budget.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-resource-budget.json
index 198578e92..610634ee8 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-resource-budget.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-resource-budget.json
@@ -23,8 +23,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/methods/training-resource-budget/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/methods/training-resource-budget/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/methods/training-resource-budget/SKILL.md",
         "src/inferrsi/data/skill_library/methods/training-resource-budget/metadata.json",
@@ -57,7 +57,7 @@
     "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/references/source-readset.md": "2a5c166767f56fa47e7eba7386f1f36da8cb73702e5b5bbe14d6a456ca9ebfae",
     "src/inferrsi/data/skill_library/methods/training-resource-budget/references/sources.md": "6d248f58af8b350fc66c22898988d6f451ac60d03bcd805cd6e84df5e55174ac",
     "src/inferrsi/data/skill_library/methods/training-resource-budget/references/stas-decision-notes.md": "f1a993e6c4ca9a035a2effcaf5500081f227b0cc56cc885a6b22fda0041fc744",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "training-resource-budget",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/weight-loading-optimization.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/weight-loading-optimization.json
index d14c9f9a8..a90165a86 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/weight-loading-optimization.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/weight-loading-optimization.json
@@ -24,8 +24,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md",
         "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json",
@@ -42,7 +42,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md": "df168c23a6f0fadddc147925a11e9eefc61f340be8e7c01a3fbe3aab002aa490",
     "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json": "c45d544312627787d523142e25730f1c3aa1b0f007ee9c10dcf0cec9f3044f71",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "weight-loading-optimization",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/diagnose-attention-bottlenecks.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/diagnose-attention-bottlenecks.json
index 6bdba8b52..fe3fbe14c 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/diagnose-attention-bottlenecks.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/diagnose-attention-bottlenecks.json
@@ -16,8 +16,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json",
@@ -52,7 +52,7 @@
     "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md": "4d4cabe0991eab554dcfbd4fa9c8199e61fa4767709b0c4cd646e2588bb398a2",
     "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md": "5ed8f888443e38ccb72545b6025447aab7b92936f78c151fed680e582126239b",
     "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md": "53436e0de4e6dda2d29426591bf61f8526378cdc6f9ec7967b88b1bcd4c61e6f",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "diagnose-attention-bottlenecks",
diff --git a/src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/host-overhead-profiling.json b/src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/host-overhead-profiling.json
index a41df1626..d802a645c 100644
--- a/src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/host-overhead-profiling.json
+++ b/src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/host-overhead-profiling.json
@@ -22,8 +22,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1/src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/SKILL.md",
-      "commit": "86c7dd8d70cbc1e47dd85c79e9fef7afb92599f1",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/32c744c841ab64ac4e106749603da9b1111b726e/src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/SKILL.md",
+      "commit": "32c744c841ab64ac4e106749603da9b1111b726e",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/metadata.json",
@@ -40,7 +40,7 @@
   "source_files": {
     "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/SKILL.md": "326e566fbe30d3ba04e4b50c3e87aad9df037f0ac4ad6fe129165ae97dc5f91d",
     "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/metadata.json": "e89f1fbbfdf63b666d42adec7826eb4b4a80428b288e476ddb9c977d9aeb3ab2",
-    "src/inferrsi/data/skill_library/skill-index.json": "8d86c12b3a8632bef0d7560aa702c67e3d41b7590873e6c336bb7714670f9939"
+    "src/inferrsi/data/skill_library/skill-index.json": "9b437dd34dbb396473f75a4f0baf11c0b4948e2259b587db2c1d00c2dca4dcee"
   },
   "skill_metadata": {
     "name": "host-overhead-profiling",
diff --git a/src/inferrsi/data/skill_library/README.md b/src/inferrsi/data/skill_library/README.md
index 423020d2f..c6718ea41 100644
--- a/src/inferrsi/data/skill_library/README.md
+++ b/src/inferrsi/data/skill_library/README.md
@@ -1,6 +1,6 @@
 # InferRSI skill library
 
-The English-only library contains **71 skill documents**: 50 registered skills
+The English-only library contains **72 skill documents**: 51 registered skills
 and 21 engine and research-router skills. Complete procedures, metadata,
 references, cases and source notices ship with the package.
 
@@ -20,7 +20,7 @@ existing resource directory.
 
 ## Read registered skills
 
-[`skill-index.json`](skill-index.json) maps 50 stable skill IDs to their canonical
+[`skill-index.json`](skill-index.json) maps 51 stable skill IDs to their canonical
 paths and backend scopes. Install the optional `skills` extra for these commands:
 
 ```bash
@@ -46,7 +46,7 @@ device compatibility; readers do not execute helpers or promote candidates.
 
 Five repository entry skills, 15 focused sub-skills and one research router
 provide source-grounded procedures. These 21 documents are available through
-the links and catalog records below, separately from the 50-ID registry.
+the links and catalog records below, separately from the 51-ID registry.
 The registered ComfyUI audit is listed alongside these imports and uses its
 generated skill abstract rather than a separate runtime repository pack.
 
@@ -105,3 +105,11 @@ separates graph construction, cross-stream ordering, backend buffer retention an
 result availability. It is source-derived candidate guidance; CPU, Metal and CUDA
 runtime behavior and performance remain unverified. MLX-LM request/cache policy and
 MLX platform discovery belong to their separate owners.
+
+## MLX-LM application guidance
+
+[Trace MLX-LM generation and cache lifecycle](repositories/mlx-lm/generation-cache-lifecycle/SKILL.md)
+links prefill/decode boundaries, three-stage batch-row ownership, prefix-cache
+identity and cooperative stopping. It is a source-derived candidate with no
+installed API, model, native-backend or service validation. MLX framework execution
+and MLX platform discovery remain separate responsibilities.
diff --git a/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md b/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md
new file mode 100644
index 000000000..33d81ec99
--- /dev/null
+++ b/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md
@@ -0,0 +1,27 @@
+---
+name: generation-cache-lifecycle
+description: Trace MLX-LM prefill, generation batching, prompt-cache reuse and cooperative stopping when request state, cache rows or retained prefixes become inconsistent.
+---
+
+# Trace MLX-LM generation and cache lifecycle
+
+Use for mixed-length batches, prefix-cache reuse, unexpected retained state, stop/removal behavior or mismatches between single generation and the server. Do not use for MLX platform discovery, generic GPU memory allocation, model training, or unverified serving-performance claims.
+
+This is a source-derived candidate with runtime validation not_run. Read the [pinned source map](references/source-map.md), [generation/batch path](references/generation-batch.md) and [cache/stop contract](references/cache-stop.md). MLX owns array execution; MLX-LM owns request identity, cache policy, batching and token termination. The two repositories have independent source pins and unverified combined compatibility.
+
+## Inputs
+
+Record MLX-LM and MLX revisions/builds, backend/device, exact model/tokenizer/adapter/draft-model bytes or revisions, entry point, tokenized prompts, cache classes, batch/prefill limits, per-request samplers/processors/stop sequences, seed and KV quantization settings. For a failure, keep a bounded sequence of request IDs, stage changes, token/cache lengths, completion reason and the next request's result.
+
+## Procedure
+
+1. **Resolve the entry path.** Distinguish `generate_step`, `stream_generate`, `BatchGenerator` and server dispatch. Server batching additionally requires the model to be batchable, no request seed, and no KV quantization; the library class does not remove those gates (G1/G5).
+2. **Follow prefill and decode.** Cache creation may defer to a model's `make_cache`; otherwise the factory chooses ordinary or rotating caches. Prefill chunks explicitly evaluate cache state. Decode schedules the next token asynchronously before yielding the current token; preserve the stream and read boundary when interpreting time or cache progress (G1/C1).
+3. **Keep rows aligned.** Track UID through pending, prompt-processing and generation stages. On completion/removal, filter cache rows, tokens, samplers, processors, lengths, stop matchers and pending outputs using the same row selection. Include `_currently_processing` when removing prompt-stage requests (G2/G3).
+4. **Check reusable-prefix identity.** Trace the server's model key and token prefix into `LRUPromptCache`. Reuse returns copies; trimming depends on the cache type. Record exact tokenization and model/adapter/draft identity rather than treating a text string or cache-file path as a sufficient key (C2/C3).
+5. **Trace stop through cleanup.** Handler cleanup sets a context flag. The batch loop observes it and removes the UID; this is cooperative application-level removal, not immediate cancellation of already queued MLX work. Keep normal EOS/length, text-stop, disconnect and shutdown paths separate. Single-request and distributed stop behavior differ (G4/G5).
+6. **Validate the relevant boundary.** Use the [validation plan](references/validation.md): mixed lengths, prompt-stage removal, generation-stage removal, survivor completion, prefix reuse/copy isolation and the next request. Do not silently fetch model weights or launch a server to satisfy a source-derived claim.
+
+## Output and stop conditions
+
+Return an entry-path decision, UID/stage/cache-row table, prefix key and cache-class assumptions, terminal/cleanup path, and exact passed/failed/not-run evidence. Stop at the first unsupported contract or missing required fixture/backend and retain a targeted verification plan. Do not infer a paged allocator, radix sharing or instantaneous prefill interruption from the existence of batching or a stop flag.
diff --git a/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/metadata.json b/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/metadata.json
new file mode 100644
index 000000000..93ecc3c31
--- /dev/null
+++ b/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/metadata.json
@@ -0,0 +1,45 @@
+{
+  "profile": "local-skill-v1",
+  "schema_version": 1,
+  "kind": "skill",
+  "origin": [
+    "external"
+  ],
+  "status": "candidate",
+  "empirical_validation": "not_run",
+  "license": "MIT",
+  "tags": [
+    "mlx-lm",
+    "generation",
+    "batching",
+    "kv-cache",
+    "cancellation"
+  ],
+  "sources": [
+    {
+      "type": "git",
+      "url": "https://github.com/ml-explore/mlx-lm",
+      "commit": "ab7f7048dc341ced2d84b1887eeb17d60b0d1b02",
+      "paths": [
+        "mlx_lm/generate.py",
+        "mlx_lm/models/cache.py",
+        "mlx_lm/server.py",
+        "tests/test_generate.py",
+        "tests/test_prompt_cache.py"
+      ],
+      "readset": "references/source-map.md"
+    }
+  ],
+  "applicability": {
+    "scope": "MLX-LM generation, batch-row and prefix-cache diagnosis; model/backend/runtime compatibility requires separate qualification."
+  },
+  "limitations": [
+    "Source-derived candidate; no local native-backend, model, service or efficacy validation.",
+    "Static content checks do not promote this candidate or establish installed API conformance."
+  ],
+  "provenance_readsets": [
+    "references/source-map.md"
+  ],
+  "references": [],
+  "consumption_mode": "packaged"
+}
diff --git a/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/cache-stop.md b/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/cache-stop.md
new file mode 100644
index 000000000..d8559f5ba
--- /dev/null
+++ b/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/cache-stop.md
@@ -0,0 +1,13 @@
+# Cache identity and stop ownership
+
+IDs refer to [source-map.md](source-map.md).
+
+`make_prompt_cache` defers to `model.make_cache` when available. Otherwise it creates per-layer KVCache, or RotatingKVCache with the configured maximum. Do not override a model's cache contract merely because a different model uses ordinary KV tensors (C1).
+
+KVCache's logical offset is distinct from capacity: storage grows in blocks, returned views end at the logical offset, and trimming reduces that offset. This does not prove that trimming returns physical memory immediately. BatchKVCache filters keys, values, offsets and left padding together, and may remove shared left padding (C1).
+
+The server resolves model/adapter/draft-model paths into its model key; prompt-cache lookup also consumes the token prefix (vision has a separate key path outside this skill's verified scope). LRUPromptCache returns deep copies for exact/shorter matches; reuse from a longer prefix needs trimmability and leaves a suffix for processing. Insertion updates count/byte accounting, removes redundant trimmable prefixes and evicts under limits (C2/C3). Path identity is not a content hash: pin actual model/tokenizer/adapter contents before comparing experiments.
+
+Handler cleanup sets `_should_stop`. In batched generation the generation loop observes the flag, shares removal UIDs where applicable and removes their stage/row state; normal token/text completion can retain cache via a different branch. Single generation checks the flag after a response and then still reaches cache insertion; its distributed stop branch raises NotImplementedError (G4/G5). Do not promise immediate cancellation during prefill or cancellation of already submitted MLX kernels from this flag.
+
+`BatchGenerator.close` restores a changed wired-memory limit after synchronizing its stream; it is not itself the same operation as removing one request or destroying all cache state. Model/cache destruction on the server generation thread is a separate shutdown path (G4).
diff --git a/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/generation-batch.md b/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/generation-batch.md
new file mode 100644
index 000000000..767aaebbd
--- /dev/null
+++ b/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/generation-batch.md
@@ -0,0 +1,18 @@
+# Generation and batch state
+
+IDs refer to [source-map.md](source-map.md).
+
+Single generation: tokenize/validate → make model-specific/default cache → chunked prefill with explicit cache evaluation → final-prompt step → schedule next decode token with async_eval → read/yield current token → EOS/length → final detokenizer segment (G1). The next token may already be queued when the current token is yielded; yielded-token count and computation submitted are not identical boundaries.
+
+BatchGenerator maintains pending sequences, a PromptProcessingBatch, a GenerationBatch and prompt-stage `_currently_processing`. `_next` performs existing decode work first; full completion capacity prevents further prompt work. Available capacity admits pending sequences, segments are consumed in bounded chunks, and prompt rows with only their final input move to generation (G2).
+
+| Per-row state | Keep aligned during filter/remove |
+|---|---|
+| Identity/data | UID, token history, cache row, pending token/logprob |
+| Generation configuration | sampler, logits processors, max token count, stop sequences |
+| Progress | generated count, token context, stop matcher |
+| Prompt processing | prompt batch and `_currently_processing` row indices |
+
+Completion extracts the finished row's cache before filtering survivors. Explicit removal searches all three stages; a caller may request caches before removal (G2/G3). No request-level state should be inferred from a single tensor's length alone.
+
+Server requests reach batching only through the server's actual gates (G5). Library batching support, a server thread and HTTP streaming are separate facts; none proves an arbitrary configuration batches or meets latency objectives.
diff --git a/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/source-map.md b/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/source-map.md
new file mode 100644
index 000000000..10014f2d8
--- /dev/null
+++ b/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/source-map.md
@@ -0,0 +1,21 @@
+# MLX-LM source map
+
+Source: [ml-explore/mlx-lm](https://github.com/ml-explore/mlx-lm/tree/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02), revision `ab7f7048dc341ced2d84b1887eeb17d60b0d1b02`. Inspected on 2026-10-21. Copyright © 2023 Apple Inc.; [MIT license](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/LICENSE). This bounded source readset does not qualify the installed API or model/backend combination.
+
+CODE is an implementation observation; TEST-SOURCE is an unexecuted upstream assertion. The diagnostic steps are HYPOTHESIS until applied to a matching workload.
+
+| ID | Class | Pinned path/symbol | Observation |
+|---|---|---|---|
+| G1 | CODE | [generate_step L388-L493](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/mlx_lm/generate.py#L388-L493); [stream_generate L707-L782](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/mlx_lm/generate.py#L707-L782) | Prefill evaluates cache state; decode uses its generation stream and async submission; stream wrapper handles tokenization/EOS/length/final text. |
+| G2 | CODE | [BatchGenerator._next/next L1866-L1967](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/mlx_lm/generate.py#L1866-L1967); [GenerationBatch.filter/next L1502-L1573](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/mlx_lm/generate.py#L1502-L1573) | Decode runs first, capacity gates prompt admission, and completion filters all per-row generation state. |
+| G3 | CODE | [BatchGenerator._find_uids/remove L1767-L1823](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/mlx_lm/generate.py#L1767-L1823) | UID lookup spans pending/prompt/generation; prompt removal also filters `_currently_processing`. |
+| G4 | CODE | [GenerationContext.stop L318-L332](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/mlx_lm/server.py#L318-L332); [batch stop/removal L1014-L1042](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/mlx_lm/server.py#L1014-L1042); [handler finally L1705-L1706](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/mlx_lm/server.py#L1705-L1706) | Context stop is a flag; batch loop propagates UID removal; generation-thread exit resets the model/cache. |
+| G5 | CODE | [server batching gate L766-L771](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/mlx_lm/server.py#L766-L771); [single path L1144-L1157](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/mlx_lm/server.py#L1144-L1157) | Seed/KV quantization constrain batching; single-request stop still reaches cache insertion and distributed stop raises. |
+| C1 | CODE | [cache factory L15-L40](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/mlx_lm/models/cache.py#L15-L40); [KVCache L339-L407](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/mlx_lm/models/cache.py#L339-L407); [BatchKVCache.filter L982-L999](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/mlx_lm/models/cache.py#L982-L999) | Model-specific cache factory, logical offset versus allocated capacity, and aligned batch padding/offset filtering are distinct contracts. |
+| C2 | CODE | [ModelProvider.load L492-L501](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/mlx_lm/server.py#L492-L501); [single lookup L1083-L1094](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/mlx_lm/server.py#L1083-L1094) | Model key is resolved model/adapter/draft tuple; lookup also uses the token/vision cache key. |
+| C3 | CODE | [LRUPromptCache.fetch/insert/trim L1652-L1732](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/mlx_lm/models/cache.py#L1652-L1732) | Exact/shorter/longer prefix paths copy cache state; longer-prefix reuse requires trimmability; byte/count constraints evict entries. |
+| T1 | TEST-SOURCE | [removal regression L773-L802](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/tests/test_generate.py#L773-L802) | Prompt-stage removal keeps currently-processing rows aligned and the survivor generates. |
+| T2 | TEST-SOURCE | [test fixture L26-L30](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/tests/test_generate.py#L26-L30); [cache fixture L29-L42](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/tests/test_prompt_cache.py#L29-L42) | Both test classes load a Hugging Face model; selecting an apparently small method still invokes setup. |
+| D1 | CODE | [pyproject dependencies/extras L11-L47](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/pyproject.toml#L11-L47); [cache_prompt command L92-L146](https://github.com/ml-explore/mlx-lm/blob/ab7f7048dc341ced2d84b1887eeb17d60b0d1b02/mlx_lm/cache_prompt.py#L92-L146) | MLX backend variants are explicit; cache command loads a model, generates state and writes a cache artifact. |
+
+Not covered: all model classes, full speculative decoding, VLM cache semantics, quantization kernels, distributed serving correctness and training. No model/native test, server, benchmark or device validation is represented by this readset.
diff --git a/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/validation.md b/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/validation.md
new file mode 100644
index 000000000..e3bca60ff
--- /dev/null
+++ b/src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/validation.md
@@ -0,0 +1,11 @@
+# Validation plan
+
+The candidate is source-derived. It has no installed API, native backend, real-model or HTTP qualification.
+
+- Inspect test class setup before selecting a native method. Both TestGenerate and TestPromptCache load `mlx-community/Qwen1.5-0.5B-Chat-4bit`; a short method selection does not avoid setup. Use an already authorized local fixture and an offline dependency configuration, or leave it not_run. Never change the test's assertions or replace its real model with a fake to claim a pass.
+- For row lifecycle, validate different prompt lengths, per-row stop/max-token settings, prompt-stage removal, generation-stage removal and a surviving request. Then issue another request and check that removed state did not shift identities.
+- For prefix reuse, compare copied cache state against a non-reused reference under identical model/tokenizer/adapter/draft identity. Exercise exact/shorter/longer keys, trimmable versus non-trimmable cache classes and count/byte eviction. Separate logical token trimming from physical-memory release.
+- For stopping, distinguish EOS, length, text stop, disconnect and shutdown; record the last token submitted and yielded, cleanup and subsequent request. A source flag is not an instantaneous prefill/kernel-abort guarantee.
+- CPU state checks may validate selected row/cache logic. Required Metal/CUDA dispatch and serving behavior need the matching actual backend, model and workload. A CPU import is not a replacement.
+
+Keep passed/failed/not-run evidence separate. Serving latency/throughput needs a separately frozen end-to-end comparison; this skill supplies no performance threshold or reproduced gain.
diff --git a/src/inferrsi/data/skill_library/skill-index.json b/src/inferrsi/data/skill_library/skill-index.json
index 9c27fb4f5..e6ba4d305 100644
--- a/src/inferrsi/data/skill_library/skill-index.json
+++ b/src/inferrsi/data/skill_library/skill-index.json
@@ -200,6 +200,10 @@
     "lazy-evaluation-stream-lifetimes": {
       "path": "frameworks/mlx/lazy-evaluation-stream-lifetimes",
       "backend": "common"
+    },
+    "generation-cache-lifecycle": {
+      "path": "repositories/mlx-lm/generation-cache-lifecycle",
+      "backend": "common"
     }
   }
 }
diff --git a/tests/check_installed.py b/tests/check_installed.py
index 1aaa63583..5e7116a6b 100644
--- a/tests/check_installed.py
+++ b/tests/check_installed.py
@@ -17,7 +17,7 @@
     "This check requires an installed package"
 )
 records = knowledge_records()
-assert len(records) == 92, "Knowledge is missing from the installation"
+assert len(records) == 93, "Knowledge is missing from the installation"
 assert (
     files("inferrsi")
     .joinpath("data", "repo_packs", "vllm_omni", "knowledge", "README.md")
@@ -27,7 +27,7 @@
 assert search_knowledge(repo="vllm_omni", backend="ascend")
 assert all(not item["verification"]["locally_verified"] for item in records)
 skill_index = search_knowledge(repo="inferrsi", limit=100)
-assert len(skill_index) == 50, "Missing installed skill abstracts"
+assert len(skill_index) == 51, "Missing installed skill abstracts"
 assert all(record["index_type"] == "packaged-skill" for record in skill_index)
 indexed = KnowledgeReader().read("inferrsi.ascend.skills.ascend_operator_profiling")
 assert "data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md" in indexed.content
@@ -59,7 +59,7 @@
         data = resource.read_bytes().replace(b"\r\n", b"\n")
         assert hashlib.sha256(data).hexdigest() == expected_hash, relative
 registry = json.loads(library.joinpath("skill-index.json").read_text(encoding="utf-8"))["skills"]
-assert len(registry) == 50
+assert len(registry) == 51
 for entry in registry.values():
     assert library.joinpath(*entry["path"].split("/"), "SKILL.md").is_file()
 router = library.joinpath("workflows", "inference-engine-research")
diff --git a/tests/test_knowledge_contract.py b/tests/test_knowledge_contract.py
index 9781724ac..2a54c8b01 100644
--- a/tests/test_knowledge_contract.py
+++ b/tests/test_knowledge_contract.py
@@ -24,7 +24,7 @@ def test_local_profiles_preserve_core_and_structured_ownership():
         assert metadata["kind"] == "skill"
         for readset in metadata["provenance_readsets"]:
             assert (sidecar.parent / readset).is_file()
-    assert len(normalized) == 45
+    assert len(normalized) == 46
 
 
 def test_contradictory_legacy_metadata_fails(tmp_path):
diff --git a/tests/test_mlx_lm_knowledge.py b/tests/test_mlx_lm_knowledge.py
new file mode 100644
index 000000000..aaae1be63
--- /dev/null
+++ b/tests/test_mlx_lm_knowledge.py
@@ -0,0 +1,33 @@
+from pathlib import Path
+
+from inferrsi.catalog import get_knowledge, search_knowledge
+from inferrsi.knowledge_operations import KnowledgeReader
+from inferrsi.skill_catalog import skill_metadata, skill_resource
+
+LIBRARY = Path(__file__).resolve().parents[1] / "src/inferrsi/data/skill_library"
+SKILL = "generation-cache-lifecycle"
+IDENTITY = "inferrsi.common.skills.generation_cache_lifecycle"
+
+
+def test_mlx_lm_candidate_preserves_source_and_application_owner():
+    metadata = skill_metadata(LIBRARY, SKILL)
+    assert metadata["status"] == "candidate"
+    assert metadata["empirical_validation"] == "not_run"
+    assert metadata["sources"][0]["commit"] == (
+        "ab7f7048dc341ced2d84b1887eeb17d60b0d1b02"
+    )
+    resource = skill_resource(LIBRARY, SKILL, "references/cache-stop.md")
+    assert resource.is_relative_to(LIBRARY / "repositories/mlx-lm" / SKILL)
+    assert "LRUPromptCache" in resource.read_text(encoding="utf-8")
+
+
+def test_mlx_lm_packaged_guidance_keeps_resources_and_unverified_state():
+    record = get_knowledge(IDENTITY)
+    assert record in search_knowledge("mlx-lm", repo="inferrsi", limit=100)
+    assert record["backend"] == "common"
+    assert record["verification"]["locally_verified"] is False
+    bundle = KnowledgeReader(skill_root=LIBRARY).read(IDENTITY)
+    assert bundle.metadata["kind"] == "knowledge"
+    for resource in ("source-map.md", "generation-batch.md", "cache-stop.md", "validation.md"):
+        assert any(name.endswith(f"references/{resource}") for name in bundle.content)
+    assert any(b"cooperative" in content for content in bundle.content.values())
diff --git a/tests/test_skill_index.py b/tests/test_skill_index.py
index bae6b1b77..cd56868dd 100644
--- a/tests/test_skill_index.py
+++ b/tests/test_skill_index.py
@@ -23,7 +23,7 @@ def test_complete_index_retains_metadata_states_and_unknown_license():
     records = search_knowledge(repo="inferrsi", limit=100)
     names = {record["name"] for record in skill_records(ROOT)}
     registry = json.loads((ROOT / "skill-index.json").read_text(encoding="utf-8"))["skills"]
-    assert len(records) == len(names) == 50
+    assert len(records) == len(names) == 51
     assert {record["skill_metadata"]["name"] for record in records} == names
     states = []
     for record in records:
@@ -41,7 +41,7 @@ def test_complete_index_retains_metadata_states_and_unknown_license():
         }
         assert set(record["source_files"]) == set(record["sources"][0]["paths"])
         states.append(metadata.get("status"))
-    assert states.count("candidate") == 45
+    assert states.count("candidate") == 46
     assert states.count(None) == 5
     profiling = next(r for r in records if r["title"] == "ascend-operator-profiling")
     assert profiling["skill_metadata"]["license"] == "UNKNOWN"
@@ -161,7 +161,7 @@ def test_package_index_and_checkout_skill_have_distinct_read_boundaries():
     assert reader.resolve("claim-contract") == "checkout:claim-contract"
     assert reader.read("claim-contract").metadata["kind"] == "skill"
     collected = reader.collect(repo="inferrsi", kind="knowledge", limit=100, material=True)
-    assert len(collected["items"]) == 50
+    assert len(collected["items"]) == 51
     assert not collected["collisions"] and not collected["omissions"]
     assert all(item["authority"] == "package" for item in collected["items"])
     assert all("record.json" in item["resources"] for item in collected["items"])
@@ -191,9 +191,9 @@ def test_index_cli_search_show_context_and_execution_boundary(capsys):
 
 
 def test_hardware_filters_include_only_matching_and_shared_skills():
-    for backend, count in (("ascend", 48), ("cuda", 37), ("cpu", 35)):
+    for backend, count in (("ascend", 49), ("cuda", 38), ("cpu", 36)):
         records = search_knowledge(repo="inferrsi", backend=backend, limit=100)
         assert len(records) == count
         assert all(record["backend"] in (backend, "common") for record in records)
     assert not (PROJECT / "skills").exists()
-    assert len(list(ROOT.rglob("SKILL.md"))) == 71
+    assert len(list(ROOT.rglob("SKILL.md"))) == 72
diff --git a/tests/test_skill_library.py b/tests/test_skill_library.py
index cc3f3fa84..5bf783f31 100644
--- a/tests/test_skill_library.py
+++ b/tests/test_skill_library.py
@@ -154,7 +154,7 @@ def test_library_root_exposes_only_owner_categories_and_required_entrypoints():
     assert {path.name for path in LIBRARY.iterdir() if path.is_dir()} == {
         "hardware", "frameworks", "repositories", "methods", "workflows",
     }
-    assert len(list(LIBRARY.rglob("SKILL.md"))) == 71
+    assert len(list(LIBRARY.rglob("SKILL.md"))) == 72
     assert not any(path.name == "zh-CN" for path in LIBRARY.rglob("*") if path.is_dir())
     references = LIBRARY / "workflows/inference-engine-research/references"
     assert (references / "import.md").is_file()