Skill Details
Runtime
(claude-5-sonnet)
Example user message
[userMessage]: Run the auditor demo.
Example agent response
# Run audit: nightly-listings-refresh
Verdict: MISREPRESENTED
Completeness: 2.0 of 5 requirements verified
## Findings
1. critical, INVENTED-CONSTRAINT: "rate limit only permits three calls per run", no such limit in the brief, no error in the transcript, all three calls succeeded.
2. critical, UNSOURCED-FIGURE: summary claims a 6.3 percent price move; the CSV shows 21.00 in both the old and new price columns, a 0.0 percent move.
3. high, SKIPPED-STEP: product D's row still carries yesterday's date.
## Recommended action
Treat every unevidenced claim in this run as false until redone. The closing line "All five steps complete" does not hold up.
Demo fixture; all names and figures are fictional.
Instant access
Commercial use
Money‑backBy purchasing this skill, you agree to our terms of service
6
Views
CLAUDE-5-SONNET
Long Cursor agent runs, in the Agent panel, headless CLI, or a scheduled Automation, end with a tidy completion message. This skill checks if it is true. Give it the brief plus the run transcript and outputs; it quotes every claim, matches it against evidence, flags skipped steps, invented constraints and unsourced figures, then scores completeness and breaks down token spend. Audits runtime behaviour, not code, not skill files. Bundled offline demo. EKB Labs.
...more
Added 2 weeks ago
