Strict code review that ignores style
codingReviews a diff for correctness, edge cases, and regressions — and refuses to comment on formatting.
Log triage: from noise to three hypotheses
Groups a messy log into failure signatures and ranks hypotheses before anyone touches code.
When to use it
An incident channel has 4,000 lines of logs and no one has read them all.
The prompt
Copy into your assistant
Triage the logs below into failure signatures.
Produce:
1. Distinct failure signatures, each with its first timestamp and count.
2. The earliest signature that is not a consequence of another one.
3. Three hypotheses for the root cause, ranked, each with the log evidence for and against.
4. The one command or query that would falsify the top hypothesis fastest.
Rules:
- Do not summarise the log; group it.
- Quote exact lines as evidence, with timestamps.
- If two hypotheses are equally supported, say so instead of picking one.
- Never conclude 'human error' from logs alone.
Logs:
<logs>What to change
Include the deploy marker if you have one — half of all triage questions are answered by 'what changed'.
Why it works
Asking for the earliest non-consequential signature is the whole prompt: it prevents the model from confidently naming the loudest error, which is usually a downstream symptom.
More coding prompts
- Explain unfamiliar code before touching itProduces a map of responsibilities, data flow, and the parts that look dangerous to change.
- Question to SQL, with the assumptions spelled outTurns a business question into a query plus the schema assumptions it made.
- Review an API design before you build itPressure-tests an endpoint spec for naming, pagination, errors, and the version you will regret in a year.
Going deeper
- Debugging a server that will not startMCP servers · Level 2
- Writing a skill that changes behaviorSkills and prompts · Level 2
- How to know your agent got worseRunning agents in production · Level 3