| name | debug |
| description | Analyze debug-analyzer.log and run tests to identify pipeline issues after a change. Standard post-change check per Captain's protocol. |
| allowed-tools | Bash Read Glob |
ultrathink
Analyze recent pipeline runs for issues.
Focus: $ARGUMENTS (leave blank for general post-change check)
For analysis-pipeline issue investigations, identify root cause before recommending fixes, prioritize quality over speed over cost, and escalate to /pipeline, /prompt-diagnosis, /report-review, or /debate only when this quick check is too narrow.
Step 1 — Read the debug log:
Read apps/web/debug-analyzer.log. If large, read the last 300 lines.
Identify: errors, warnings, stage failures, contract violations, unexpected LLM outputs, timing anomalies.
Step 2 — Run tests:
Run npm test 2>&1 | tail -40 and check pass/fail summary.
(Note: npm test is safe — it excludes expensive LLM tests.)
Step 3 — Pattern analysis:
Look for:
- Stage failures or aborts (which stage, which input)
- Evidence quality issues — unusually low probativeValue rates, claimDirection inversions
- Boundary clustering anomalies — single boundary when multiple expected, or vice versa
- Verdict direction errors — counter-evidence treated as supporting evidence
- Contract validation failures — prompt output rejected by schema validator
Step 4 — Correlate with recent changes:
Run git log --oneline -5 to see what changed.
Identify whether issues correlate with specific commits.
Step 5 — Failed-attempt recovery discipline:
If the first focused validation for the slice failed (for example npm test, npm -w apps/web run build, or the manual verification explicitly described for the change), do not jump straight to a broader fix package.
First classify the earlier attempt as keep, quarantine, or revert.
Keep only the parts still justified by the evidence from Steps 1-4.
Broaden scope only if the failure evidence shows why a wider change is required.
Step 6 — Report findings:
Summarize in priority order. For each issue: stage where it occurs, likely cause, recommended fix.
Flag any regressions in the Captain's priority benchmark families:
- Bolsonaro judgment (question and statement variants)
- Hydrogen vs. Electric Cars
- Venezuela article
If regressions are found, recommend running /validate for a full measurement.
