resolve-checks_skill

Overview

This skill automates a disciplined workflow to resolve all failing CI checks on the current branch's pull request by running tests locally, fixing failures, and ensuring CI turns green. It emphasizes running the full test suite locally first, diagnosing flakiness, and iterating until the PR is safe to merge. The goal is stable, reproducible fixes—not repeatedly re-running CI until luck wins.

How this skill works

Identify the current branch and associated PR, then run the project test sequence locally: type checks and linting, backend tests, frontend tests, RLS tests, integration tests, and behavior tests. When failures appear, reproduce the failing test, fix the root cause, re-run locally (including repeated runs for flaky tests), push the commit, and monitor CI until all checks pass. For CI-only failures, compare local and CI environments, inspect logs, and address environment differences or timing issues.

When to use it

• When CI checks are failing on your pull request
• Before attempting to merge a PR to ensure a stable merge
• After making code changes and before pushing to avoid red CI
• When you suspect flaky tests or environment-specific failures
• When a reviewer requests that all checks be green

Best practices

• Run tests sequentially: check -> backend -> frontend -> rls -> integration -> behavior
• Reproduce a failing test locally before making changes; run the single test file
• Diagnose flakiness by running the test many times and fixing root causes (timing, race, non-determinism)
• Compare local and CI environments (runtime versions, env vars, parallelism) for CI-only failures
• Commit clear messages describing the fix and why it resolves the failing checks

Example use cases

• A unit test fails after a refactor; run the unit suite locally, fix logic, and push the fix
• A DB test breaks due to a schema change; update fixtures and re-run db tests before pushing
• A test passes locally but fails in CI due to timing; identify race condition and add proper awaits or polling
• An integration test fails in CI because of missing credentials; detect and document the environment difference
• A flaky test intermittently fails; stabilize it by removing non-determinism and validating with repeated runs