triage-issue
Investigate a bug, regression, or failing test without immediately implementing a fix. Use when the task is to reproduce an issue, isolate the failure surface, identify likely root cause, and recommen
Files1
View on GitHubSKILL.md.claude/skills/triage-issue/SKILL.md
GitHubTriage Issue
Use this skill when the immediate goal is diagnosis, not implementation.
Rules
- Reproduce before theorising whenever feasible
- Prefer the narrowest failing case available
- Separate confirmed facts from hypotheses
- Do not edit code unless the user explicitly asks to move from triage into fixing
- End with a concrete recommendation for next action
Step 1 — Define the Symptom
Establish:
- Expected behavior
- Actual behavior
- Trigger conditions
- Scope of impact
If a failing test, stack trace, or error text exists, use it as the primary entry point.
Step 2 — Reproduce and Narrow
Prefer one or more of:
rg -n "<error text|symbol|route|function>"
<targeted test or build command>
git diff --stat
Determine:
- Whether the failure is reproducible
- Which component owns the failing behavior
- Whether the issue appears new, pre-existing, or caused by recent changes
Step 3 — Form the Diagnosis
Summarise:
- Confirmed failure surface
- Most likely root cause
- Confidence level
- Unknowns still blocking a definitive conclusion
If multiple causes remain plausible, rank them instead of forcing certainty.
Output
Return:
- Reproduction status
- Suspected root cause
- Evidence
- Affected files or modules
- Recommended next step
Done Criteria
The triage is complete only when:
- The symptom was scoped clearly
- Reproduction was attempted when feasible
- The likely root cause was narrowed to a concrete area
- The final response separates evidence from speculation