| name | result-audit |
| description | Use when an operator must review outputs for unsupported conclusions, weak evidence, missing ablations, broken assumptions, or incomplete reporting. |
Result Audit
When to Use This Skill
Use this skill when the node needs to:
- review a report, notebook, or artifact bundle
- detect unsupported conclusions
- check whether the evidence matches the claimed contribution
- flag missing baselines, controls, or ablations
- judge whether the final output is ready for handoff or publication
Workflow
- Read the stated claim before the evidence.
- Map each important claim to the supporting artifact or result.
- Check for missing comparisons, controls, or caveats.
- Identify where the output overstates certainty.
- Return findings as prioritized review items.
Expected Outputs
- a severity-ranked review list
- unsupported claim flags
- missing evidence or baseline flags
- final release recommendation
Quality Rules
- focus on evidence, not writing style
- do not accept "looks plausible" as support
- call out missing baselines explicitly
- separate hard failures from optional improvements
