fix(autopilot): install action-library plugins in CI + PAT retry re-label - #59
Merged
Conversation
…retry re-label claude-code-action does not read the target repo's .claude/settings.json, so the executor, gates, and metrics workflows now install the content-stack plugins explicitly via plugin_marketplaces/plugins inputs. The retry step re-adds autopilot:run with AUTOPILOT_PAT — a GITHUB_TOKEN label event never re-triggers the executor, so retries were silently inert. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…lable Private repos on the GitHub Free plan cannot enable allow_auto_merge or branch protection. The auto-merge job already runs behind all four gates, so falling back to a direct squash merge is equivalent there; SKILL.md documents the plan limitation and its trust implications. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Found live on the sandbox: the action fails during setup with 'Could not fetch an OIDC token' without it, even when authenticating via anthropic_api_key. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- build job: setup-node + corepack (pnpm was exit 127 on a bare runner) - executor action gets github_token: AUTOPILOT_PAT so task PRs are authored by the PAT identity, not claude[bot] - gate actions allow the claude bot actor as fallback Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Both gates completed their reviews on the live sandbox run but exceeded --max-turns 20 (26 and 21 turns), failing the action on principle. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The job checks out with fetch-depth: 0; re-fetching base --depth=1 clobbers origin/<base> and loses the merge base once base advances, failing the three-dot diff with 'no merge base'. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- audit/anti-slop gates post their P0/Must Fix findings as PR comments before writing the sentinel (red gates were silent — findings lived only in the unlogged Claude final message) - new report-failure job comments on the PR and the task issue when any gate fails, closing the triage loop - executor self-reviews content with /content-ops:review-content before opening the PR so articles don't systematically bounce off the gate Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
… whole-repo Found live: a pre-existing P0 (site missing in astro.config) blocked a clean content PR. The repo must pass /astro-builder:audit before the loop is activated. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…resets on unblock The executor prompt orders plugin-skill invocations but Skill was not in --allowedTools, so the run died on a permission denial. Retry counting now only considers failure markers newer than the last autopilot:blocked un-label, so a human unblock grants a fresh retry budget instead of stale markers sending the task straight back to blocked. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
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
1 participant
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.
Two executor-path fixes found while preparing milestone 2:
claude-code-action@v1doesn't read the target repo's.claude/settings.json; the executor/gates/metrics prompts invoke/astro-builder:audit,/content-ops:review-content, etc., which would all fail. Each invocation now passesplugin_marketplaces+pluginsexplicitly (executor gets all three library plugins; each gate gets only what it uses).autopilot:runwith the defaultGITHUB_TOKEN, whose label events never trigger workflows (same anti-recursion rule as the P0 fixed in feat(autopilot): init skill — scaffold the loop's control-plane surface #57). It now usesAUTOPILOT_PAT.🤖 Generated with Claude Code