| name | triage-bug |
| description | Investigate a bug or unexpected behavior by tracing the root cause through the codebase, then produce a TDD-based fix plan. Use this whenever a user reports a bug, crash, or wrong behavior — even if they say "something is off", "this broke", "not working", "why is X happening", "investigate this", or paste an error message. Also use when the user asks to trace a regression, understand a failure, or plan a fix. This is a hands-off deep-dive that ends with a tracked issue and a concrete fix plan. |
Triage with TDD
Investigate a reported problem, find its root cause, and produce a TDD fix plan. The goal is to hand the user a clear, actionable issue they can immediately start working on. Minimize back-and-forth — code exploration answers most questions faster than asking.
Process
1. Capture the problem
If the user hasn't described the issue, ask exactly one question: "What's the problem you're seeing?"
Resist asking follow-ups before you investigate. You'll find most of the context yourself, and what you can't find you can clarify at the end.
2. Explore and diagnose
Launch parallel explore subagents to investigate different angles simultaneously — this saves significant time on larger codebases. A good split:
- Subagent A — Code path: Trace from the symptom back to its origin. Follow the call chain, data flow, or event lifecycle until you find where the behavior diverges from what's expected.
- Subagent B — History & tests: Check git log on relevant files for recent changes. Look at existing tests to understand what's already verified and what's missing.
- Subagent C (if needed) — Similar patterns: Search the codebase for analogous features that work correctly. This reveals whether the issue is unique or systemic.
For small, focused bugs a single subagent is fine — use your judgment on scope.
Your investigation should surface:
- Where the bug manifests (entry point, UI, API, data layer)
- What code path is involved
- Why it fails — the root cause, not just the symptom
- Whether it's a regression, missing guard, design flaw, or missing feature
3. Assess and synthesize
Summarize what you found:
- Root cause (be specific about the mechanism, not just the location)
- Confidence level: high / medium / low — be honest if investigation was incomplete
- Type: regression | missing feature | design flaw | edge case | environmental
- Minimal fix scope: which modules or contracts need to change
Keep descriptions at the module/behavior level. Avoid file paths and line numbers — the issue should stay useful even after the code is reorganized.
4. Design TDD fix plan
Create an ordered list of RED-GREEN cycles. Each cycle is one vertical slice of the fix:
- RED: Describe a specific test that captures the broken or missing behavior
- GREEN: Describe the minimal code change that makes that test pass
Good cycles:
- Test through public interfaces, not implementation internals — so they survive refactors
- Assert on observable outcomes (API responses, UI state, return values), not internal state
- Each cycle stands alone — implement one before writing the next
Aim for 2–5 cycles. One is a sign the problem is too narrow; more than five suggests the scope should be split.
End with a REFACTOR note if any cleanup would be valuable once the tests are green.
5. File the issue
Call the report-issue skill with the full triage context: root cause, fix type, TDD plan, and acceptance criteria. The report-issue skill will handle formatting and submission to the appropriate tracking system.
Output
Once the issue is created, print:
Issue: <URL>
Root cause: <one sentence>
Fix: <N> RED-GREEN cycles — first test: <brief description of RED #1>
