[audit-workflows] Agentic Workflow Audit — 2026-10-22

[github-actions[bot]]

Summary

The sampled runs show recurring driver exits and a high AI Moderator failure rate, with zero missing tools. The 30-day request returned 338 runs from only 17 hours on October 22; the 24-hour request returned 100 runs from 4 hours 44 minutes. Both hit collection limits, so these are partial samples rather than full-period rollups.

Collection and health

• 24-hour request: 100 records, 97 executed, 86 succeeded, 11 failed; raw and adjusted success rate 88.7% (no intentional-failure runs in this sample). There were 18 errors, 10 driver exits, 1 agent-logic failure, 0 missing tools, and 0 warnings.
• 30-day request: 338 records sampled from 2026-10-22 03:48–21:18 UTC; 325 executed. Raw success was 278/325 (85.5%). Excluding the one Daily Credit Limit Test run gives 277/324 (85.5%). The sample had 82 errors, 37 driver exits, 9 agent-logic failures, 0 missing tools, and 0 warnings.
• Engine counts from the logs summary: agy 1, aider 3, claude 32, codex 54, copilot 129, pi 105 (324 classified runs).

Findings and recommendations

1. Triage shared driver exits. Handle agent failure appeared in 32 failures across 17 workflows; Checkout PR branch appeared in 20 failures across five workflows. Both failure patterns recur in recent audits. Review the common conclusion handler and checkout step across affected workflows.
2. Investigate AI Moderator failures. The 30-day request summary reports 8 failures across 10 runs (80%). This hotspot also appeared in the October 17 audit history.
3. Repair log pagination and continue the audit window. Both requested ranges stopped at the count limit, leaving the 24-hour and 30-day coverage incomplete. The trend charts below therefore show only the returned October 22 sample.
4. Review the token outlier. Daily Cache Strategy Analyzer used 3,377,391 tokens in 35 action minutes, the largest sampled run.

Trend charts

The collected October 22 bucket contains 277 successful and 47 failed executed runs after excluding the intentional-failure run, for an adjusted 85.5% rate. Only one daily bucket was available, so this does not show a 30-day trend.

The sampled bucket contains 53,668,233 tokens after excluding the intentional-failure run; the largest individual run used 3.38 million tokens. A seven-day moving average is unavailable because the fetch produced only one daily bucket.

View full findings

• Recurring finding IDs: driver_exit_failure_hotspot, ai_moderator_failure_hotspot, and incomplete_log_coverage.
• 24-hour sample window: 2026-10-22 16:35:40–21:19:20 UTC.
• 30-day request sample window: 2026-10-22 03:48:33–21:18:35 UTC.
• 30-day sample aggregate: 53,714,423 tokens before excluding the intentional-failure run; 53,668,233 after exclusion.
• Coverage anomalies: count limit reached for both requested windows; episode membership is partial.

References:

• §39829351871 — PR Code Quality Reviewer failure
• §39797577184 — AI Moderator failure
• §39832668788 — Daily Cache Strategy Analyzer token outlier

Generated by 🔍 Agentic Workflow Audit Agent · codex · gpt60 · ~4 AIC · ⌖ ~1 AIC · ⊞ 16.2K · ◷

• expires on Oct 23, 2026, 1:39 PM UTC-08:00

[github-actions[bot]]

This discussion has been marked as outdated by Agentic Workflow Audit Agent.

A newer discussion is available at Discussion #71497.