Flaky test hunter
Finds tests that fail intermittently in CI and proposes fixes.
RunsEvery Monday at 9 AM
Cron
0 9 * * 1Works withAny agent, any stack
Look at the last 50 CI runs on the main branch with gh run list and gh run view --log. Identify tests that both passed and failed on the same commit or with no related code change. For each flaky test, read the test code, explain the likely cause (timing, shared state, network, ordering), and propose a concrete fix. Output a ranked list by failure frequency.
Replace anything in <ANGLE BRACKETS> with your own details. Review outputs before letting an agent send, buy or delete anything.
How to run this
Works with any agent
⏱TriggerEvery Monday at 9 AM
✦Any agentRuns the prompt
✓OutputResult for you to review
Steps
- Copy the prompt and fill in the parts in angle brackets.
- Paste it into whatever agent you already use: a chat assistant, a coding agent, an automation tool, or a script that calls any model API.
- Schedule it (every monday at 9 am). Use the agent's own scheduler if it has one, otherwise run it from cron with the expression below.
- Give it only the access it needs (inbox, calendar, repo, sheet) and keep sends, purchases and deletes behind your approval.
- Review the first two runs, tweak the prompt, then let it run.