Repository navigation
Conversation
Claude-Session-Id: 01a15cb6-ff42-75a0-8c3a-712b63716008
Gavel-Issue-Id: 9723fa02-92e0-4ae0-ae6f-1815602e2cba Claude-Session-Id: 01a15ea7-403c-7d71-854f-08688a7af676
Claude-Session-Id: 01a15e89-a154-70d0-beff-58fa5d2ba7b1
Claude-Session-Id: 01a161f3-946a-7f80-8de1-6443c175cde5
Add the Captain Formation brand kit: horizontal, stacked, symbol, and wordmark masters, dark and one-colour variants, a small-size drawing, and web and social icons. Lettering is outlined vector geometry, so no font installation is required. Web handover files are provided for manual integration; application code is unchanged. PNG exports follow the existing *.png ignore rule. Trademark clearance and print approval remain outstanding.
… review prompts Presets gain an integer order (default 0) that sorts layers ahead of scope precedence, so the Lint preset can apply after prompt and request layers. The Lint preset replaces output schemas with a violations contract that gavel lint prompt renders as lint results. Five read-only document review prompts ship in pkg/ai/prompt/builtin, and local prompts with the same name shadow them. Adds permissions explain and permissions convert commands. JSON/YAML null on schema fields now clears them explicitly. With saved defaults, a missing token budget stays unset so the runtime applies its own default.
Add a portable access level (--access read-only|full, permissions.access) that compiles per runtime into native posture: tool denies on claude, a read-only sandbox that never asks on codex, and plan mode on gemini. Add `captain permissions explain` to show the compiled posture per runtime, and `captain permissions convert` to rewrite a permissions block into the portable form, refusing any conversion that would loosen access. Ship the document review prompts as a read-only built-in source. A user copy in the config prompts directory shadows a built-in of the same name instead of making the name ambiguous. Expose --preset, --no-presets, --runtime-profile and --vars on prompt commands, and route prompt runs through a shared PromptInvocation so the CLI and HTTP paths resolve prompts the same way. BREAKING CHANGE: an unset max-tokens no longer defaults to a built-in 4096. It is left to the runtime's own default (0 = runtime default), so output length can change for prompts with no saved or frontmatter value.
Ship reviewer prompts that Captain resolves by name through the new prompt.Builtin embed FS, so users can lint internal and public high-level and low-level specs and RFP proposals without copying prompt files into their repositories. Each prompt is read-only, runs with a 30-turn budget, and reports violations against a fixed rule catalog. Shared rule ids (req/, doc/) keep findings consistent across document types, and audience-specific ids cover internal and public concerns.
Authors can state read-only workspace access once with permissions.access. CompileAccess lowers it to each runtime's native dialect: per-tool denies on claude, a read-only sandbox with approval never on codex, and the plan approval mode on gemini. Layers resolve in order through ResolveAccessState, so a later layer can lift a read-only intent to full. A runtime that cannot reach the posture is refused rather than silently loosened, for example codex with the sandbox off or looser raw cmux flags. Fallback candidates are compiled and validated as they will run, so an unreachable fallback fails before the primary does. ConvertToAccess rewrites dialect-specific denies or sandboxes into the portable form only when the result is equivalent or tighter. Fallback substitution keeps the original runtime before crossing providers, so an agent is not moved to API mode and left without its tools. SpecLayer gains an Order field that sorts before scope. The YAML marshaller no longer re-emits raw opaque fields. BREAKING CHANGE: saved-spec resolution no longer defaults budget.maxTokens to 4096. The runtime now chooses the value, so callers that relied on 4096 must set it explicitly.
… worktree workspaces Wrap every provider candidate, primary and fallbacks, in an access decorator so permissions.access read-only is compiled for the runtime that candidate actually runs on. A fallback in another family now receives its own dialect instead of the primary's. Explicitly supplied directories that are linked worktrees are recorded in the run workspace before the first turn, so active runs show their workspace while running. Detached HEAD is recorded as no branch. Active runs started before this change are reconciled without altering their execution state. Runtime presets gain a sort_order column and a Lint built-in preset is added. Config home resolution moves to captainconfig.Home.
Codex execpolicy rules run matching commands outside the sandbox, even under a read-only sandbox with approval never. A user rule such as prefix_rule(["git","add"], decision="allow") could let a read-only run write the git index. `codex exec` can skip rules with --ignore-rules, but the app-server and TUI cannot. Read-only runs for those modes now start in a captain-managed CODEX_HOME under captain's config directory. It links only auth.json, config.toml and AGENTS.md from the user's home, so it carries no rules. Runs refuse a project .codex/rules layer that allows commands, because the isolated home cannot hide project rules. Session discovery now also scans the isolated home so its rollouts remain visible. BREAKING CHANGE: read-only runs fail when project execpolicy rules allow commands, and when the isolated home contains a rules directory.
…to the run's thread Execpolicy allow rules could run commands outside the sandbox even under read-only, so read-only runs now start in an isolated CODEX_HOME and pass --ignore-rules. A codex process cannot change CODEX_HOME, so a session refuses a turn that needs a different home. Notifications from sub-agent threads are dropped so they no longer replace the run's session id or end its turn.
…ead-only runs Claude's tool denies do not follow sub-agents, teams, peer messages, crons or worktrees the agent creates, so read-only runs could still write through them. Those tools are now denied, while read-only listing tools stay allowed. Codex execpolicy allow rules run commands outside the read-only sandbox. `codex exec` now passes --ignore-rules, and the app-server and cmux TUI start in an isolated CODEX_HOME without user rules. BREAKING CHANGE: read-only codex runs are refused when project rules in the tree allow a command, and read-only Claude runs lose the newly listed tools.
Append the prompt source search paths to human-readable output of prompt list, so users can see where prompts are loaded from. JSON, YAML, and file exports are unchanged. Replace the shared help for prompt subcommands with command-specific help that includes descriptions, examples, and usage. The root prompt help gains a Help and output section covering color, NO_COLOR, --json, and agent-session Markdown. The help text now states that permissions.tools 'ask' is not enforced by native tool filters and that per-call approval goes through the broker. The sample prompt's Edit tool policy changes from ask to allow to match this.
Read-only codex runs write rollouts under captain's isolated CODEX_HOME, which the monitor did not watch or accept, so those transcripts were never ingested. Watch day directories and resolve threads across both codex session roots. Accept transcript paths under any provider root, and validate ingest paths against all of them.
Bring the docs site in line with the current prompt runtime and CLI. - Add a Prompt CLI page covering discovery, authoring, render/run/observe, and help and color behavior. - Document portable permissions.access (read-only/full), how it compiles per runtime, and the codex execpolicy handling. - Replace runtimeProfile with presets and --preset, and add the Lint built-in preset. - Update approvals for the broker's OnRunState hook and Deadline, and the read-only escalation rules. - Add reference pages for Sandboxes, Configuration, Serve, Sessions, and Fixtures, and move them out of the placeholder section. - Rewrite the Gavel TODO docs around the PostgreSQL-backed lifecycle, durable plans, and fixture verification. - Correct the prompt API path to /api/v1/prompt and the prompt schema examples to typed variables and model/mode fields. BREAKING CHANGE: runtimeProfile and --runtime-profile are deprecated and ignored; use presets in frontmatter or --preset instead.
Shorten the root README to an overview, quick start, prompt authoring, verification, and a command map. Detailed reference material (fixtures, sandbox credentials, git-agent deployment, whoami and configure details) moves to the docs site. Update the runtimeprofiles README to the preset model. RequestedProfile, PinnedProfile, DefaultProfile, the prompt runtimeProfile field and --runtime-profile are deprecated in favor of RequestedPresets, frontmatter presets and --preset. Mark the unified approval callback plans as historical design records that point to the current approvals guide. Fix the DeepSeek attachment example to pass the prompt with -p, matching the current CLI.
Move the promptfile lint command into the rootcmd package; main.go referenced a symbol from a sibling file, breaking the standalone entrypoint spec. Also discard the unused String() result flagged by go vet in the version spec.
|
Important Review skippedToo many files! This PR contains 547 files, which is 447 over the limit of 100. To get a review, reduce the PR to 100 files or fewer by splitting it into smaller PRs or changing its base branch. Upgrade to a paid plan to raise the limit. Usage-priced reviews support at most 300 files.
|
Gavel summary
Totals: 0 passed · 0 failed · 0 skipped · - |
|
Review the following changes in direct dependencies. Learn more about Socket for GitHub.
|
Gavel summary
Totals: 6769 passed · 20 failed · 12 skipped · 6m13s Failing testsruntimeprofiles — Default runtime catalog > uses supplied config and deduplicates configured directory aliasesmonitor — API transcripts > refuses an API transcript path outside Captain's transcript directorycli — AIRuntimeOptions --perms > fails on an unknown perm and lists the available namescli — prompt help > honors plain help output > root flagcli — prompt help > honors plain help output > root environment... and 15 more failing tests — see the full gavel-results artifact. |
… catalog Expose native provider configuration, reusable resource definitions and Captain portable and backend configuration alongside the execution schemas. Each catalog document gains a stable id and kind. Execution modes and existing URLs are unchanged. Add a stateless POST /api/captain/ai/adapter-schemas/validate endpoint. It compiles local references only and returns instancePath and schemaPath errors without resolving, executing or persisting anything. Generated OpenAI agent request types move to OpenAIAgentOptions* names, and union fields keep their JSON marshaling. Form annotations (help links, enum labels and icons, typed JSON editors) are generated from pinned sources by hack/adapterresources. BREAKING CHANGE: GET adapter-schemas now also returns resource documents. Clients must key on id and kind rather than assuming a provider/mode pair. Go types ThreadStart, ThreadResume and TurnStart are renamed to OpenAIAgentOptionsThreadStart, OpenAIAgentOptionsThreadResume and OpenAIAgentOptionsTurnStart. Claude-Session-Id: b00416a9-facd-49f8-9bc6-cd4ef7eba6b4
…w UI Add POST /api/captain/ai/adapter-schemas/validate to check a preview value against an embedded schema without executing or saving it. The schema page gains form, JSON, and YAML views with local drafts, resource search, and provenance links. Catalog entries now include id and kind, and mode is optional for portable resources. BREAKING CHANGE: unknown provider/mode routes under /adapter-schemas now show an error instead of falling back to the first schema. Claude-Session-Id: b00416a9-facd-49f8-9bc6-cd4ef7eba6b4
Prompts can now declare a `tools` mapping in frontmatter. Each tool runs its `exec` script with the tool input as JSON on stdin and returns stdout to the model, or parsed JSON when an output schema is set. Captain previously rejected dotprompt's list of registered tool names, so prompts had no way to ship their own tools. Scripts run through the commons-db shell runner and inherit the run's setup (dotenv, env vars, connections), with per-tool values layered on top. Timeouts kill the whole process group. Retry supports backoff and exit-code filters. The readonly and destructive hints set the default permission, and toolPolicy still governs tools by name. Inline tools are validated at parse and lint time: names, exec, hint conflicts, schemas, and a warning for retried destructive tools. Runs that would bypass the runner (chat launch, supplied providers) refuse them instead of silently dropping them. Also included: symlinked prompt directories in local sources, self-contained draft renders, and CLI completion for fallback models, presets, sandboxes, and permission modes. Claude-Session-Id: b00416a9-facd-49f8-9bc6-cd4ef7eba6b4
… launch API Chat turns (mode "chat") settle their runs without ending their sessions: the chat outlives a turn and awaits the next message, so migration 84 keeps its sessions running and idle between turns. The chat writes their terminal lifecycle itself when it ends (cancelled if stopped, failed on error, otherwise as its last run ended), so a follower can hold one stream open across turns. Add LaunchPrompt, AwaitPromptRun and StopPromptRun so Go callers can start an async prompt run, await its summary and stop it through the same broker the HTTP handler uses. PromptInvocation.Tools exposes caller tools to the agent on every runtime; runtimes that cannot expose them (CLI-only, chat, observe) are rejected instead of silently dropping them. Local prompt sources now resolve read-only symlinked prompt files and immediate linked directories, reporting broken links and cycles. Transcript messages are projected through the shared session.ConversationalMessages in ingestion, session read and live follow, so all three assign the same ids and a follow resumes after the read's newest message. BREAKING CHANGE: a chat's sessions no longer go terminal after each turn; they stay running and idle until the chat ends, so clients that treated a settled chat turn as session completion must wait for the chat's end. Claude-Session-Id: b00416a9-facd-49f8-9bc6-cd4ef7eba6b4
… SDK folds them into one result Previously a turn counted prompts and expected one SDK result per prompt. The SDK folds prompts steered mid-turn into the running turn and reports them together in user_message_uuids, so a single result can settle several prompts and the count never reached zero. Each prompt now carries a uuid and is tracked until a result names it. The turn ends on the idle that follows once every prompt it sent is settled. A steer the bridge rejects is withdrawn, so it cannot hold the turn open. An SDK query-loop error aborts the turn immediately. Bridge protocol is bumped to v5 (prompt uuid, consumed uuids on turn/completed). The logging middleware also logs rejected replies at debug level, since parse errors name only the first bad byte. Claude-Session-Id: b00416a9-facd-49f8-9bc6-cd4ef7eba6b4
…t rendering Document prompt lookup locations in root and prompt list help, and render unsaved drafts without a saved id. Linked prompts are now read-only: update, delete, and create-in-place refuse them and symlinked destinations. Caller tools are appended to every runtime variant, failing fast on runtimes that cannot expose them. `prompt run --vars` derives its schema from the prompt's input schema and defaults. BREAKING CHANGE: updating or deleting a prompt reached through a symlink now fails instead of writing through it. Claude-Session-Id: b00416a9-facd-49f8-9bc6-cd4ef7eba6b4
Claude-Session-Id: b00416a9-facd-49f8-9bc6-cd4ef7eba6b4
Add generated JSON schemas describing the native configuration for gemini-cli 0.64.0 (google.json) and codex-cli 0.164.0 (openai.json). The adapters can now expose and validate provider-native settings with field types, enums, defaults, and documentation links. Each schema records its source URL and SHA-256, so drift from upstream can be detected. The files are generated by hack/adapterresources and should be regenerated rather than edited by hand. No existing behavior changes. Claude-Session-Id: b00416a9-facd-49f8-9bc6-cd4ef7eba6b4
…tificate authority Move the egress credential proxy from pkg/gitagent/proxy to pkg/credproxy and serve CONNECT. HTTPS clients tunnel through CONNECT, which was refused, so their requests never reached grant substitution. With an Authority configured, the proxy terminates tunnels to granted https destinations using a per-run CA and runs each inner request through the same grant pipeline. Ungranted destinations and tunnelled requests whose Host differs from the CONNECT target are still refused. The kubeproxy package moves from pkg/ai/fixture to pkg/credproxy and switches from Director to Rewrite. BREAKING CHANGE: the package path pkg/gitagent/proxy is removed in favour of pkg/credproxy. CONNECT is now served when an Authority is set, so clients using tunnelled HTTPS must trust the Authority certificate.
…nline tools Inline tools always ran on the host with the run's full credentials, even in sandboxed runs, while the docs said sandboxed runs confined them. Each tool now declares where it runs (placement: host|sandbox) and how credentials reach it (credentials: none|direct|proxied). When credentials is unset it is derived from the run's setup, and a sandbox placement that would expose secrets is refused until a mode is chosen. Proxied tools get a credential-free kubeconfig and an egress proxy that injects grant headers and enforces per-destination scope. HTTPS goes through CONNECT with a per-run CA whose key stays in the Captain process. Tool environments are now prepared once per run and torn down at PhaseRun. Credential files go to a private temp directory instead of the workspace cwd. Unsupported combinations, such as placement: sandbox on off or native runs, and any inline tool under docker or git-agent, fail during preflight before the model is called. Existing prompts keep today's behavior by default. kubeproxy and gitagent/proxy move to pkg/credproxy.
…and sidecar placement Git-agent runs refused inline tools because the remote agent had no route back to them. The sidecar already relays results to the supervisor over SSH or HTTPS, so the supervisor now publishes its caller tools and per-tool credential services over that same transport. Each service opens only for the task's dispatched agent and is withdrawn once the outcome arrives. Tools declare placement: supervisor (the default, run on the submitting host) or sidecar (run beside the worktree). Two new credential modes, sidecar and sidecar-proxied, let a sidecar tool use its own identity or receive grant values once into an in-sidecar proxy. Proxies now require a per-tool secret, and the runner scrubs the agent's environment when it holds credentials. The agent and runner share an OS user, so the scrub and deny-read rules keep secrets out of the agent's environment and context. They do not stop a same-user process that escapes confinement. Codex rejects the denied-read policy, so credentialed sidecar tools fail on Codex agents. BREAKING CHANGE: RemoteExecutor.Execute takes RemoteExecOptions. ToolPlacement.Effective and Sandbox.ValidateToolPlacement are replaced by ResolveToolPlacement. InlineTool.ResolveCredentials takes the placement. kubeproxy.Start takes Options, and WriteKubeconfig is now a package function. Claude-Session-Id: 90b00fac-abd5-497e-a270-84918c0b3cce
…summary Broken symlinks in discovered or explicit prompt paths used to fail the whole run. They now surface as promptfile/symlink warnings, so healthy prompts still lint and the run exits successfully. Unreadable regular files still fail as before. Human output gains an error/warning summary line. The JSON output stays a plain array. Also exposes BindInlineTools and SetRun for chat hosts, and routes npm install output through the logger instead of stderr.
Broken or unreadable prompt links now appear as warnings in listings instead of failing the whole list. Render and run still fail for the selected prompt. Name lookup uses the first matching local source before embedded ones, so duplicate basenames no longer error as ambiguous. Log output is routed through the event renderer and committed above the streaming turn, including while an approval form is open. Chat sessions bind inline tools as the runner does instead of refusing prompts that declare them. Shared runtime flags are grouped into help categories (Model runtime, Model permissions, Model selection). BREAKING CHANGE: duplicate prompt names in local sources no longer return an ambiguity error; the first matching source wins.
| } | ||
|
|
||
| func readServices(repo, task string) (Services, error) { | ||
| data, err := os.ReadFile(servicesPath(repo, task)) |
Tool loops were hard-capped at maxToolTurns (16), ignoring the caller's budget. The request's budget.maxTurns now takes precedence, then the provider config's, then the 16-turn default. Runs that set neither value behave as before.
… embed The reusable prompts were embedded from pkg/ai/prompt/builtin and exposed through the exported Builtin embed.FS. They now live under examples/ as plain files, and the Builtin variable is removed from pkg/ai/prompt/examples.go, so the prompts are no longer resolvable by name from Go code. The lint runtime preset also changes schemaStrictness from error to repair, so schema violations in lint output are repaired instead of failing the run. BREAKING CHANGE: prompt.Builtin is removed. Callers that resolved built-in prompts by name must load them from examples/ or from their own source instead.
…ailure Make 'repair' the canonical schema strictness; 'retry' remains an alias. When a response is rejected, its Markdown text is shown as a diagnostic event and kept as the response text if repair fails. Cost and usage from every attempt are now accumulated rather than dropped, and the loop returns partial results on error. Anthropic models are registered in the genkit plugin catalog, and Markdown detection moves to ai.IsMarkdown. BREAKING CHANGE: provider.IsMarkdown is removed in favor of ai.IsMarkdown. The 'terra' alias now resolves to gpt-6.1-terra instead of gpt-5.6-terra.
…er task groups Rejected schema output was discarded, so callers lost the response text and its spend. Rejected text now streams ahead of the terminal error, and both run paths return the output with the error. Launched runs can join a caller's task group under its concurrency, and session overviews fold tokens, cost and tool calls from the transcript until the monitor projects them. BREAKING CHANGE: the built-in prompt source is removed. Review prompts now live under examples/ and must be found through a prompts directory.
…up precedence Built-in review prompts now live in examples/ and are no longer embedded, so the docs direct users to copy them locally. Describe schema repair, chat tool binding, symlink diagnostics, and bare-name lookup order to match current behavior. BREAKING CHANGE: spec-internal-high, proposal-review and the other review prompts no longer resolve by name from the binary; copy them from examples/ into a prompt directory.
Gavel-Issue-Id: 31bf3fff-ecb6-418b-aecc-7a8f5cb9d60e Claude-Session-Id: 01a17ba1-5e5b-7bd3-add8-a0a38f1804ce
Add key:=json, key[]=@file or key[]=- list values, and key[sub]=value nested keys to --var. Keep each --var whole (split:false, CSV decode) so JSON and commas in values are not split. Add a join helper so templates can render list variables with a separator. Examples and docs now pass files as a typed list. BREAKING CHANGE: the bundled review prompts expect files to be a list; use files:=["..."] instead of files=...
Switch the files input from a delimited string to a non-empty array and render it with join, so paths are no longer split on spaces or commas. BREAKING CHANGE: files must now be passed as a JSON array (e.g. --var 'files:=["a.md","b.md"]' or files[]=@list.txt); space- or comma-separated strings are no longer accepted.
…pecs Add spec.batch to split a run's input files into runs of N files, optionally executed concurrently. Add HookAction for schema and input hooks and for workflow.after steps that run a tool or child prompt, gated by CEL 'if' predicates. Replace the commit gates enum with a list of named workspace gates. The retired none, cheap and full strings still parse. Record caller tools and tool policies as ignored in spec provenance for runtimes that take no tools (jev). BREAKING CHANGE: api.CommitGates, AllCommitGates and Commit.EffectiveGates are removed. Commit gates are now a list of named workspace gates, and Commit.Gates no longer takes a single enum value.
Every commit is now cut through the commons-db workspace pipeline (stage, hooks, gates, commit). Content cloned from the source's gitignored files never enters a commit or keeps a worktree alive. The built-in git committer and the fixed cheap gates (secret file names, 1 MiB size cap) are replaced by per-policy workspace gates, and a gate runs only when the policy lists it. HostMessage, HostGates and HostHooks let a host compose messages and add checks without bypassing the pipeline. Do remains only as a temporary gavel compatibility seam. The same change adds OpenTelemetry spans for agent runs and model calls, tool and prompt actions for workflow.after steps with if/with expressions, and input/output hooks for TypeSafe prompts. The jev provider now accepts caller tools instead of rejecting them. BREAKING CHANGE: gates: cheap, full and none and Hook.MaxFileSize are removed in favor of named gate specs. A misspelled gate name now fails the commit. anchor: auto is refused rather than delegated to a host callback.
…acing to prompt runs
Bind --otel-collector, start the tracer in PersistentPreRunE (a configured collector that fails to start is a hard stop) and flush it after Execute so spans from failed runs are delivered. Promote otel, otel/sdk and otel/trace to direct requires, bump commons to v1.63.0, and attach the prompt input-file flags the batching commit added. Claude-Session-Id: 1c483396-b792-4048-a3ae-756864511b40
…tmatter Input files passed after a prompt are now split into batches with --batch and --batch-concurrency, and each batch is recorded as its own run. Hook actions (tool, child prompt or commands) can run at input, schema and after-run points, under the tool's permission. Templated frontmatter is parsed as declared and keeps its authored text for reserialization. Jev now ignores inline tools instead of refusing them, leaving them to hooks, so Jev runs that declare tools no longer fail preflight. Child prompts started by a hook are filed as child runs via parent_run_id. BREAKING CHANGE: Jev runs that declare tools no longer fail preflight; the model never receives those tools. promptcontext.Source is now an alias for promptlib.HookSource, and renderFrontmatter is exported as RenderFrontmatter.
…n key Prompt specs can now declare input.hooks, which replace the run's variables before the body renders, and output.hooks, which replace the TypeSafe or JSON Schema before validation. Both dispatch tool or prompt actions through the new hookaction package, and plan hook discovery now uses promptlib.DiscoverHookFiles. An admission session now holds at most one transcript child, enforced by a partial unique index. Migration 04 demotes extra transcript children left by earlier runs, and migration 82 relabels only the oldest Gavel agent child so it cannot create a second transcript. Push-side commit gates now come from the workflow's configured gate specs over the pushed range instead of a hardcoded secret-name check. Snapshots skip paths the workspace excludes. BREAKING CHANGE: commit gates no longer run unless a commit policy configures them.
…hout resuming them Adds a deliver channel for sending a message into a running claude or codex session that captain did not launch. Delivery uses the agent's own channel: Claude Code's messaging socket, the codex queue, or typing into its cmux pane. A session with no live process returns 409 and one with no reachable channel returns 422, so captain never resumes a session in a second process.
…s it to Claude creates CLAUDE_CODE_MESSAGING_SOCKET after it starts and exports it only to its children, so reading it off the claude process never found it: delivery always fell back to typing into the cmux pane. Take it from a direct child stamped CLAUDE_PID=<claude pid> (cmux's hooks claude inbox-wait helper), and ignore the socket and session id a nested claude inherited from its parent. Claude-Session-Id: 9d25c110-1df9-4e72-a8b3-9f07ffae70b0
… environment Delivery failed when a claude child had exited but was not yet reaped, because reading its environment returned an error. The snapshot now skips children with no environment left to read, while other unreadable children still fail the read. This replaces the per-pid child listing and the syscall.Kill ESRCH probe with one process snapshot.
…e mode field Claude-Session-Id: e9049061-cd29-4ed0-8fa5-f0472da6f62d
Enable WithObservers in DefaultOptions and wrap the providers used in schema repair execution and streaming repair with observingProvider. Repair attempts are now observed the same way as primary calls, so they appear in observer output instead of being invisible. Claude-Session-Id: ffcff737-fce3-40c5-92c8-f31bd3eef60d
…l events Add a process-wide observer registry to the AI middleware so callers can receive request, response, stream, and agent iteration and tool events from provider calls. Observers are optional and unobserved calls stay allocation-free. Environment values in published requests are redacted, and the agent runner publishes its iteration and tool events through the same channel. The change is additive and does not alter existing behavior. Claude-Session-Id: ffcff737-fce3-40c5-92c8-f31bd3eef60d
…des are discovered The workspace `.captain/prompts` directory was only appended when no writable source existed, so a configured source directory silently shadowed local overrides. It is now always included once. Add exported save/remove helpers so frontends can persist prompt edits without duplicating validation, base-version checks, or deletion logic. Refresh the toolchain and dependency graph (Go 1.27.1, clicky, commons-db, gavel 0.0.67, OpenTelemetry 1.47) and bump the gavel action, flanksource/deps, and the base image digest to keep CI and container builds reproducible. Claude-Session-Id: 01a19aaf-b85f-7c10-bd27-e36b616fe72a
What
permissions.access(read-only|full), compiled per runtime, withcaptain permissions explainandconvert--preset,--no-presets,--runtime-profile,--vars, and a sharedPromptInvocationCODEX_HOMEand--ignore-rules; read-only Claude runs deny out-of-process toolscaptain lintfor promptfiles, DeepSeek Flash image attachments, and per-command prompt helpassets/brand/main.gois self-contained sogo run ./main.gocompilesWhy
Notes
max-tokensno longer defaults to 4096;0means the runtime default.runtimeProfileand--runtime-profileare deprecated and ignored; use presets.*.pngignore rule. Trademark clearance and print approval are still outstanding.