| name | execution-artifact-triage |
| description | Triage saved runtime artifacts before editing code. Use when the user asks to inspect project/execution, project/errors, console.md, terminal.md, output.md, saved logs, warning output, large log-like files, or asks whether a warning/error needs a fix. |
Execution Artifact Triage
Use this skill when saved runtime evidence is the source of truth.
Workflow
-
Read artifacts first.
- Inspect the named files before code edits.
- Check timestamps, repeated sections, stale noise, truncation, and whether artifacts are tracked evidence.
- If the user asks about large logs, measure size and references before recommending deletion.
-
Classify evidence.
- Current blocker, current warning, benign framework noise, stale prior failure, upstream/service issue, environment issue, or generated output artifact.
- Separate saved console/terminal evidence from assumptions about the runtime.
-
Trace only as needed.
- Use code search to map the logged route/component/export/provider path.
- Avoid changing code just to silence a warning unless it affects behavior, tests, build, packaging, or user output.
-
Recommend or proceed.
- If the user asked only to investigate, provide cause, impact, and next action.
- If the user asked to fix, hand off to
workflow-bug-fix-plan-and-logs with the classified evidence.
- Prefer truncating or preserving tracked evidence logs over deleting referenced artifacts.
Guardrails
- Do not treat a non-empty log as current truth without checking recency and context.
- Do not delete tracked evidence files unless explicitly asked and references are reconciled.
- Do not conflate old npm audit/provider failures with the current warning under investigation.