Illustrative output. All data below is fictional — used to show format, depth, and buyer confidence level.
| # | Severity | Pattern | Evidence | Recommended Fix |
|---|---|---|---|---|
| F-01 | HIGH | Silent completion without artifact write | Agent log shows "Task complete" but no output file at expected path reports/daily.json. Task repeated 3x across 2 weeks. |
Add artifact existence check after each completion signal. If path missing after T+5min, trigger retry or alert. |
| F-02 | MEDIUM | False positive health signal | Health endpoint returns 200 OK even when downstream API key is expired. 4-day outage went undetected because health checks were green. | Add synthetic health-check probe that exercises a real read/write cycle. Health ping should fail if downstream auth fails. |
| F-03 | LOW | Stale blocker accumulation | Blocker log shows 12 items flagged "needs review" older than 72 hours. None were ever resolved or dismissed — just accumulated. | Implement a weekly stale-blocker review. Blockers unresolved after 72 hours should escalate to owner notification automatically. |
| F-04 | MEDIUM | Output drift from expected schema | Agent outputs changed schema on 2026-04-20 — date field switched from ISO 8601 to UNIX timestamp without version bump or migration note. | Add schema version tagging to all agent outputs. Include a schema changelog in the artifact manifest. Lock output schemas after deployment. |
| F-05 | HIGH | Context window exhaustion without fallback | Long-running agent task (14-hour pipeline) silently dropped inputs after context window limit. Final 6 steps processed with truncated context — outputs are incomplete but not flagged. | Add context usage monitoring. If available window < estimated required context, trigger summarization or chunking BEFORE proceeding. |
$149 flat — 24-hour turnaround — results or full refund.
Buy AI Ops Checkup — $149