Repository navigation
[Copilot Dreaming] Remove false guarantees and duplicate CI advice; locate the right checks and host - #283
Open
Copilot wants to merge 8 commits into
Open
[Copilot Dreaming] Remove false guarantees and duplicate CI advice; locate the right checks and host#283Copilot wants to merge 8 commits into
Copilot wants to merge 8 commits into
Conversation
… rules Copilot could mistake the available test choices for a two-way split. The instructions describe internal Go tests, command tests and catalog scenarios, so an extra fixed count adds a misleading taxonomy to the placement decision. https://github.com/github/gh-actions-lock/blob/90ae563599cf303ff91eb5b46277ff04f5d2b4fc/AGENTS.md#L43-L48 https://github.com/github/gh-actions-lock/blob/90ae563599cf303ff91eb5b46277ff04f5d2b4fc/AGENTS.md#L51-L54 The cheapest-sufficient-layer rule and prohibition on repeating a claim still apply across the choices. Each layer's coverage limits remain in its bullet. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Copilot could look for a job named test when checking command-test coverage. The owning workflow defines go-test and runs make test there, so naming that job makes the check discoverable without changing its test or authentication requirements. https://github.com/github/gh-actions-lock/blob/90ae563599cf303ff91eb5b46277ff04f5d2b4fc/.github/workflows/test.yml#L10-L28 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…a host Copilot could troubleshoot the remote's host even when GH_REPO selects another repository host. Host resolution delegates its repository tier after the explicit hostname and GH_HOST overrides. https://github.com/github/gh-actions-lock/blob/90ae563599cf303ff91eb5b46277ff04f5d2b4fc/cmd/gh-actions-lock/root.go#L341-L353 The hostname test exercises a host-qualified GH_REPO, establishing why that existing override belongs in the selection guidance. https://github.com/github/gh-actions-lock/blob/90ae563599cf303ff91eb5b46277ff04f5d2b4fc/cmd/gh-actions-lock/verify_test.go#L19-L55 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
The same skill reader already receives the verify-in-CI action in the mode table, together with its network, write and blocking behavior. Repeating the invocation below adds another place for maintainers to update without changing the reader's decision. https://github.com/github/gh-actions-lock/blob/90ae563599cf303ff91eb5b46277ff04f5d2b4fc/skills/gh-actions-lock/SKILL.md#L34-L40 CI verification remains prescribed in that table. The adjacent warning that offline coverage does not establish trust remains in place. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Copilot could treat an IO-failed fix as an unchanged checkout. Workflow writes precede later action-file writes and lockfile saving, and a later error returns without restoring earlier writes. https://github.com/github/gh-actions-lock/blob/90ae563599cf303ff91eb5b46277ff04f5d2b4fc/internal/pin/commit.go#L59-L85 https://github.com/github/gh-actions-lock/blob/90ae563599cf303ff91eb5b46277ff04f5d2b4fc/cmd/gh-actions-lock/run.go#L529-L534 Removing the broad guarantee avoids that assumption. The distinct no-write gate for diagnosed blocking findings still applies before commit, and the instruction to commit workflow edits with the lockfile remains. https://github.com/github/gh-actions-lock/blob/90ae563599cf303ff91eb5b46277ff04f5d2b4fc/cmd/gh-actions-lock/run.go#L508-L523 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…ntee Copilot could reject valid JSON by requiring every documented finding key. The serializer omits several fields when empty or false, so a field vocabulary is useful but a universal presence promise is not. https://github.com/github/gh-actions-lock/blob/90ae563599cf303ff91eb5b46277ff04f5d2b4fc/cmd/gh-actions-lock/format/json.go#L59-L73 All field names remain available to readers; no output field or schema compatibility rule changes. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Copilot could rely on periodic verification after repository activity stops. The recommended workflow runs on pull requests and pushes to main, with no scheduled event. https://github.com/github/gh-actions-lock/blob/90ae563599cf303ff91eb5b46277ff04f5d2b4fc/.github/workflows/verify-lockfile.yml#L6-L25 Removing the weekly promise makes the copied workflow's coverage clear. Its actual events, copy-then-lock order and token-access warning remain unchanged. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Copilot could expect an attachment that a failed invocation never produced. Offline runs return before log registration, and JSON failures do not print the saved path. https://github.com/github/gh-actions-lock/blob/90ae563599cf303ff91eb5b46277ff04f5d2b4fc/cmd/gh-actions-lock/run.go#L139-L164 Saving is also best-effort, so even an online failure need not leave a log. https://github.com/github/gh-actions-lock/blob/90ae563599cf303ff91eb5b46277ff04f5d2b4fc/cmd/gh-actions-lock/runlog.go#L63-L85 The attachment duty still applies to saved run logs. Removing the unconditional creation claim avoids confusing a missing current-run artifact with a separate defect. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
There was a problem hiding this comment.
🟢 Approval recommended
No active issues were identified in the review, with the documentation updates focused on removing inaccurate guidance and reflecting the existing CLI and CI behavior.
No open issues 🎉
What changed in this PR
This documentation-only PR aligns Copilot guidance with the actual CLI behavior and CI configuration, removing unsupported guarantees and duplicate advice.
Changes:
- Documents
GH_REPOin host-selection guidance and identifies the actualgo-testCI job. - Removes inaccurate guarantees about atomic writes, JSON fields, saved logs, and scheduled workflow runs.
- Simplifies overlapping verification guidance.
| File | Description |
|---|---|
| skills/gh-actions-lock/SKILL.md | Updates host-selection, verification, JSON, run-log, and CI guidance to match the implemented behavior. |
| AGENTS.md | Removes the misleading test-layer count and names the actual go-test CI job. |
Review effort: Lite (auto)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
🧠 How this helps Copilot
Copilot could mistake the test-placement choices for a two-way split or look for a Go test job that does not exist. Removing the false count and naming the actual check makes it easier to choose and find the right coverage.
Host troubleshooting can target the wrong service when
GH_REPOis set. Including that override makes the selection guidance match the resolver. The CI mode table already supplies the verification command, so removing its second mandate also avoids maintaining the same advice twice.An IO-failed fix can leave earlier file writes behind. Removing the unconditional all-or-nothing promise avoids encouraging Copilot to treat every failed checkout as unchanged.
JSON findings can omit optional keys, and failed invocations need not produce or print a saved log. Removing those universal promises helps Copilot consume valid reports and handle unavailable attachments without losing the field vocabulary or saved-log attachment duty.
The copied verification workflow has no weekly trigger. Removing that promise avoids relying on checks that will not run without repository activity.
Review commit by commit for each change's rationale and evidence.
Note
Copilot Dreaming curates this repository's Copilot instructions and skills in the background, proposing updates to help agents work here. Learn more about Copilot Dreaming.
Automated details
Source:
curate-memoriesDream, analyzingmainat90ae563.Discarded guidance: A tag-selection source comment remains unchanged because source-comment edits are outside this run's instruction and skill scope. The earlier profiling addition is not carried forward because this run did not establish an eligible new profiling lesson; existing profiling help remains unchanged.