Headless Developer Pipeline: AI-assisted end-to-end lifecycle

[TheSethRose]

Summary

Implement a headless AI-assisted developer pipeline that handles the full task lifecycle through GitHub, from context ingestion to merge and cleanup.

Current behavior

• Worker scripts exist for task grooming and execution in skills/task-manager/scripts (backlog-groomer.js, recurring-worker.js, review-worker.js).
• Cron setup exists in docs/CRON-SETUP.md with worker orchestration patterns.
• GitHub integration is available for issue sync and repo metadata in src/lib/github.ts and project sync routes.

Target behavior

• A single, end-to-end workflow that can:
  • Ingest task context from the issue and comments.
  • Create and link a feature branch tied to the issue.
  • Execute autonomous changes with staged commits and conventional messages.
  • Open a pull request with an AI-generated summary and testing notes.
  • Track CI checks and report status updates.
  • Merge on approval and clean up branches.

Subtasks (all included in this issue)

1. Context ingestion
   • Fetch issue title/body/comments and map into a structured task brief.
   • Store brief in work_notes for auditability.
2. Isolated workspace
   • Create a feature branch per issue with a deterministic naming scheme.
   • Record branch name on the task or issue to prevent duplication.
3. Autonomous development
   • Run edits in the project repo only (honor local_path/workspace_path).
   • Stage and commit with conventional commit messages.
   • Add progress notes via add-work-note.js.
4. Automated submission
   • Create a PR via gh CLI with summary, scope, and test steps.
   • Link the PR back to the issue.
5. CI/CD validation
   • Poll GitHub checks via gh or API until success or failure.
   • Post status updates to work_notes.
6. Lifecycle resolution
   • Merge PR on approval and passing checks.
   • Close issue and delete branch.

Scope

• Pipeline controller script or service, aligned with existing worker scripts.
• Integration with gh CLI for branch/PR actions.
• Safe retry and idempotency markers in work_notes.
• Security guardrails: only act within the linked repo and branch.

Acceptance criteria

• A single command can execute the full lifecycle for one issue.
• Each stage records status in work_notes for traceability.
• Branch creation, PR creation, CI monitoring, and merge are all automated with guardrails.
• Re-running the pipeline does not duplicate PRs or branches.

Open questions

• Preferred location for pipeline controller (scripts/ vs. skills/task-manager/scripts/)?
• Required CI pass list or allowlist for merge?

References

• skills/task-manager/scripts/recurring-worker.js
• skills/task-manager/scripts/backlog-groomer.js
• skills/task-manager/scripts/review-worker.js
• docs/CRON-SETUP.md
• development/worker-workflow.md
• src/lib/github.ts
• src/app/api/projects/[id]/sync/route.ts