ci-fix-with-memory

When to invoke (auto-trigger)

Invoke automatically whenever:

• The user mentions CI is failing, red, or blocked
• gh pr checks shows a failure
• A push triggers a failed workflow run

Do NOT invoke for: local test failures (use test-fix instead), lint-only failures with an obvious single-line fix.

Self-Healing CI Pipeline with Handoff Memory

When to Use This Skill

Use this skill when:

• CI checks are failing on a PR
• You want to avoid repeating previously failed fix attempts
• You need systematic CI debugging with context preservation
• Multiple CI check types are failing (lint, tests, build, etc.)

Critical Setup

This skill requires a .handoff/ directory structure:

.handoff/
├── latest.md          # Most recent session handoff
└── known-issues.md    # Persistent log of failed approaches

If these don't exist, the skill will create them on first run.

Pipeline Steps

Phase 1: Ingest Context (Before ANY Action)

1. Read the handoff file

   cat .handoff/latest.md
   

   This contains:

   • What was tried in the previous session
   • What worked and what didn't
   • Current branch state
   • Open questions or blockers

2. Read the known issues log

   cat .handoff/known-issues.md
   

   This contains:

   • Approaches that have been tried and failed
   • Specific commands or fixes that don't work
   • Patterns to avoid
   • Environmental quirks

3. Acknowledge context Present a brief summary:

   ## Context Loaded
   
   ### From Previous Session
   [Key points from latest.md]
   
   ### Known Failed Approaches
   [List from known-issues.md]
   
   ### Starting Fresh Investigation
   

Phase 2: Get Current CI Status

1. Fetch current CI check status

   gh pr checks [PR_NUMBER]
   

2. For each failing check, get full logs

   gh pr checks [PR_NUMBER] --watch
   # OR
   gh api repos/{owner}/{repo}/commits/{sha}/check-runs
   

3. Create TodoWrite checklist

   - [ ] Fix lint failures
   - [ ] Fix TypeScript errors
   - [ ] Fix test failures
   - [ ] Fix build errors