---

description: Check CI pipeline status for current branch and fix failures using systematic debugging allowed-tools: Bash(gh ), Bash(git ), Bash(pnpm typecheck), Bash(pnpm lint), Bash(pnpm test*), Bash(pnpm build*), Read, Glob, Grep, Edit, Write#

Fix Pipeline

Pipeline Context#

<current_branch> !git branch --show-current </current_branch>

<push_status> !git status -sb | head -1 </push_status>

<pipeline_runs> !gh run list --branch "$(git branch --show-current)" --limit 3 2>&1 </pipeline_runs>

<latest_run> !gh run list --branch "$(git branch --show-current)" --limit 1 --json databaseId,status,conclusion,workflowName,headSha 2>&1 </latest_run>

<recent_commits> !git log --oneline -5 </recent_commits>

<changes_vs_main> !git diff main...HEAD --stat 2>/dev/null || echo "Cannot diff against main" </changes_vs_main>

---

Decision Tree#

Evaluate latest_run above:

Condition	Action
No runs / “not found”	Branch not pushed? Offer to push with git push -u origin HEAD
status: "in_progress"	Pipeline running. Wait or re-check later.
conclusion: "success"	Pipeline passed. Report success and stop.
conclusion: "failure"	Pipeline failed. Continue to Debugging Protocol below.

---

Debugging Protocol#

THE IRON LAW: NO FIXES WITHOUT ROOT CAUSE INVESTIGATION FIRST#

Skipping phases guarantees wasted time and broken fixes.

Phase 1: Gather Failure Evidence (REQUIRED)#

1.1 Get the failed logs:

gh run view <RUN_ID> --log-failed

1.2 Answer these questions BEFORE proceeding:

• Which job failed? (typecheck / lint / test / build)
• What is the exact error message?
• Which file(s) and line(s) are referenced?
• Is this a new failure or pre-existing?

1.3 Reproduce locally:

CI Job	Local Command
typecheck	pnpm typecheck
lint	pnpm lint
test	pnpm test:e2e:with-server or specific feature
build	pnpm build

If it passes locally but fails in CI, investigate environment differences (node version, dependencies, etc.)

Phase 2: Root Cause Analysis (REQUIRED)#

DO NOT propose fixes yet.

1. Read the full error - not just the first line
2. Trace to source - what code/change caused this?
3. Understand WHY - not just WHERE

Form a hypothesis: “The error occurs because [specific reason], caused by [specific change].”

Phase 3: Minimal Fix#

Only after completing Phase 2:

1. Make ONE targeted change addressing root cause
2. Verify locally using same command from Phase 1.3
3. Confirm error is gone

Phase 4: Commit and Re-run#

# Stage and commit the fix
git add -A
git commit -m "fix: [description of what was fixed]"

# Push to trigger new CI run
git push

# Watch the new run
gh run watch

---

Red Flags — STOP and Return to Phase 1#

If you catch yourself:

• Proposing fixes without reading full error logs
• Saying “let me try X” without explaining WHY
• Changing multiple unrelated things
• Skipping local reproduction
• Not understanding the error message

STOP. You are guessing. Return to Phase 1.

Escape Hatch#

If after 3 fix attempts the issue persists:

• This may be an architectural or environmental problem
• Stop fixing symptoms
• Report findings and ask user for guidance