fix-ci

name	fix-ci
description	Diagnose and fix ALL failing CI checks on the current PR in a single pass. Collects all failures first, then prescribes fixes. Use when CI is failing, tests are broken, GitHub Actions report errors, or PR checks show red.
disable-model-invocation	true
allowed-tools	["Bash(gh *)","Bash(git *)","Bash(npm *)","Bash(mvn *)","Read","Grep","Glob"]
metadata	{"source":"internal","version":"2.0.0"}

Fix CI

Diagnose and fix ALL failing CI checks in ONE pass. Never push until all known issues are resolved.

1. Ensure All Checks Are Complete

PAGER=cat gh pr view --json number,statusCheckRollup --jq '{
  pr: .number,
  failed: [.statusCheckRollup[] | select(.conclusion == "FAILURE") | .name],
  pending: [.statusCheckRollup[] | select(.status == "IN_PROGRESS" or .status == "QUEUED") | .name]
}'

If there are pending checks, wait for all checks to complete before diagnosing. Fixing 2 of 5 failures wastes a push cycle.

2. Get ALL Failed Job IDs At Once

RUN_ID=$(PAGER=cat gh run list --branch $(git branch --show-current) --limit 1 --json databaseId,conclusion --jq '[.[] | select(.conclusion == "failure")][0].databaseId')
echo "Run ID: $RUN_ID"

Then get all failed jobs:

PAGER=cat gh api repos/{owner}/{repo}/actions/runs/$RUN_ID/jobs --jq '[.jobs[] | select(.conclusion == "failure" and (.name | test("CI Status|all-ci") | not)) | {id, name}]'

3. Get ALL Job Logs At Once

For EACH failed job ID from step 2, get logs in a single loop:

for JOB_ID in <SPACE_SEPARATED_JOB_IDS>; do
  echo "=== JOB $JOB_ID ==="
  PAGER=cat gh api repos/{owner}/{repo}/actions/jobs/$JOB_ID/logs 2>&1 | tail -80
  echo ""
done

Read ALL output before making any fixes.

4. Classify ALL Failures

Before fixing anything, categorize every failure into this table. Fix in this order (earlier fixes often resolve later issues):

Priority	Category	Symptoms	Fix Command
1	Formatting	"Formatting failed", biome/prettier diff	pnpm run format
2	Lint	Biome lint errors	pnpm run check:fix
3	TypeScript	TS2xxx errors, type mismatch	Fix the type error in source
4	Build failure	Compilation errors, missing exports	Fix imports/exports, verify with pnpm run build:webapp
5	Webapp tests	"FAIL" in webapp test output	Fix test or source, verify with pnpm run test:webapp
5	App server tests	Maven test failures, assertion errors	Fix test or source, verify with cd server && ./mvnw test -Dsurefire.includedGroups="unit" -Dmaven.test.skip=false -T 2C --batch-mode -q
6	OpenAPI sync	"OpenAPI out of sync"	pnpm run generate:api
6	DB schema	"Schema drift detected"	pnpm run db:draft-changelog
6	DB ERD	"ERD outdated"	pnpm run db:generate-erd-docs

5. Fix ALL Issues

Work through the entire list. Fix root causes, not symptoms.

IMPORTANT: Do NOT push after fixing only one failure if multiple exist. Fix everything first.

6. Validate Locally Before Pushing

After ALL fixes are applied, run local validation:

pnpm run format
pnpm run check

Then run tests for ALL components that had failures:

# If webapp tests failed:
pnpm run test:webapp

# If app-server tests failed:
cd server && ./mvnw test -Dsurefire.includedGroups="unit" -Dmaven.test.skip=false -T 2C --batch-mode -q && cd ../..

ALL must pass locally before pushing.

7. Regenerate If Needed

If any OpenAPI or DB validation failed:

pnpm run generate:api
pnpm run db:generate-erd-docs

Run format + check again after regeneration:

pnpm run format
pnpm run check

8. Commit and Push (ONCE)

git add -A
git commit -m "fix(<scope>): resolve ci failures"
git push

9. Monitor

PAGER=cat gh pr checks $(PAGER=cat gh pr view --json number -q .number) --watch

Rules

1. NEVER push after fixing only one failure if multiple failures exist
2. ALWAYS run local validation (format + check + affected tests) before pushing
3. If step 1 shows pending checks, wait for all checks to complete first
4. If the same CI check fails twice in a row, investigate deeper - do not retry the same approach