From output-auditor
Multi-angle review of responses before delivery. Audits from 4 perspectives -- factual accuracy, logical consistency, completeness, and safety. Use for high-stakes responses where errors have real consequences.
How this skill is triggered — by the user, by Claude, or both
Slash command
/output-auditor:output-auditorThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Review every response from 4 independent angles before delivery.
Review every response from 4 independent angles before delivery.
Is every claim true?
Check:
Does the response contradict itself?
Check:
Is anything missing?
Check:
Could this cause harm?
Check:
CRITICAL: Wrong information that would cause harm (wrong security advice, data-destroying command) -> Must fix before delivering. If you can't fix, do not deliver.
WARNING: Incomplete or potentially misleading information -> Fix if possible. If not, add a clear caveat.
MINOR: Style issues, missing context that isn't essential -> Fix if quick. Otherwise note for follow-up.
Before sending any response with code or technical advice:
[ ] Every file path verified with Glob [ ] Every function/variable verified with Grep or Read [ ] Steps are in correct order [ ] No self-contradictions [ ] The actual question is answered [ ] Dangerous operations have warnings [ ] Rollback path exists for destructive changes
npx claudepluginhub a-ariff/ariff-claude-plugins --plugin output-auditorAudits AI-generated or prototype code for structural flaws, fragility, production risks, and maintainability issues. Surfaces hidden technical debt with severity-rated recommendations.
Challenges agent's recent claims and forces genuine re-examination using code-reading tools when users push back with 'wtf are you thinking', 'that's wrong', or similar.
Cross-examines non-trivial decisions via a fresh-context adversarial reviewer before they stand. Use when correctness matters more than speed, in unfamiliar code, or for high-stakes operations.