Add automated bugfix pipeline for unattended execution#5
Open
Add automated bugfix pipeline for unattended execution#5
Conversation
Add skills/automated.md — a controller variant that chains diagnose, fix, test, and review phases sequentially with self-correction loops, designed for CI/CD bots and automated systems that run without human interaction. - Reuses existing phase skills without modification - Configurable artifact directory, lint command, and retry limits - Escalation replaces "ask the user" for unattended mode - Updates SKILL.md and guidelines.md to reference both controllers Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…decay reminder Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The implementation notes capture design decisions, alternatives considered, and test rationale — context that a future AI session needs when addressing PR review comments without memory of the original session. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add /feedback phase for addressing PR review comments with full context recovery across sessions. The skill gathers review comments from multiple sources (task file, PR URL via gh CLI, or user input), reads prior session artifacts, implements changes, and appends a feedback round section to session-context.md — creating a running decision log that prevents context decay across multiple review rounds. The automated pipeline gains Phase 7 (Session Context), which writes a manifest summarizing key design decisions, test strategy, and known concerns from the initial session. This file serves as the entry point for feedback.md, giving the feedback session the "why" behind the original implementation. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…maries Step 7 writes a JSON mapping of comment IDs to brief summaries of what was done (or intentionally not done) for each review comment. The bot consumes this file to post descriptive PR replies instead of generic "Addressed in <sha>" messages. Gracefully skipped when comment IDs are unavailable (e.g., user-provided feedback without structured headers). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
feedback.md:
- Add Configuration section with ARTIFACT_DIR setting and path
convention table bridging interactive and automated artifact naming
- Use {pr_number} placeholder to disambiguate from issue {number}
- Add explicit comment ID extraction guidance per input source
- Add lint discovery guidance matching automated pipeline's approach
- Replace controller redirect with "stop and wait for instructions"
to prevent interactive mode drift in bot-driven sessions
- Use ARTIFACT_DIR/ prefixes throughout instead of hardcoded paths
- Add note about ARTIFACT_DIR typically being .ai-bot/ after
automated pipeline sessions
automated.md:
- Add explicit confidence threshold override for Phase 2 when
Phase 1 proceeded past low confidence
- Expand test file examples to multi-language (Go, Python, JS/TS)
- Replace loose lint auto-detection with numbered priority order
- Add Known Issues section to PR description when MAX_REVIEW_ROUNDS
exhausted
controller.md:
- Simplify phase 9 description to match other phases' style
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
adalton
added a commit
that referenced
this pull request
Mar 19, 2026
…context, and safety features Unified approach combining PR #5 (feedback skill, cross-session context) with PR #6 (flattened artifact paths, unattended foundation): - Add /feedback skill for addressing PR review comments across sessions, with dual gh api endpoint coverage, cross-session context recovery, and per-comment response tracking via comment-responses.json - Enrich unattended pipeline: lint auto-detection phase, session context manifest for future AI handoff, unit test gate check with retry counting, structured escalation (hard stop) vs graceful degradation (retry exhaustion), and configurable max_retries - Align phase numbering to 0-based (Phase 0: Start through Phase 9: Feedback) across README, SKILL.md, controller, and guidelines - Add escalation cross-reference in guidelines with note distinguishing retry-limit degradation from hard-stop escalation - Add escape hatch for unit test mandate when no test framework exists - Add lint-not-found fallback (skip and note in completion report) - Clarify retry scope asymmetry between /test and /review failure loops Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
5 tasks
adalton
added a commit
that referenced
this pull request
Mar 19, 2026
Introduces two new workflow capabilities and simplifies artifact organization across all existing skills. Supersedes PR #5 and PR #6. Unattended pipeline (/unattended): - Autonomous bugfix workflow running diagnose → fix → test → review → lint → document → session context without human interaction - Lint auto-detection with priority chain (AGENTS.md, Makefile, package.json, language defaults) - Unit test gate check counting toward max_retries - Feedback loops with configurable retry limits and graceful degradation - Structured escalation model: retry exhaustion degrades gracefully, guidelines.md criteria (security, breaking changes) trigger hard stops - Session context manifest for cross-session AI handoff Feedback skill (/feedback): - Address PR review comments across sessions with no shared memory - Recover context from prior session artifacts (session-context.md, implementation-notes.md) - Dual gh api endpoint coverage (line-level + review-level comments) - Per-comment response tracking via comment-responses.json - Declined-suggestion rationale to prevent cross-session re-litigation Artifact path flattening: - Simplify artifact paths from nested per-phase directories to flat filenames under .artifacts/bugfix/{issue}/ - Update all existing skills to use the flattened convention Cross-cutting: - Align phase numbering to 0-based (Phase 0: Start through Phase 9: Feedback) across README, SKILL.md, controller, and guidelines - Add escalation cross-reference in guidelines distinguishing retry degradation from hard-stop escalation - Add escape hatch for unit test mandate when no test framework exists
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.
Add skills/automated.md — a controller variant that chains diagnose, fix, test, and review phases sequentially with self-correction loops, designed for CI/CD bots and automated systems that run without human interaction.