[detection-analysis] Detection Analysis Report — 2026-11-07

[github-actions[bot]]

Summary

• Window evaluated: last 24 full hours (UTC), 2026-11-06T23:29:52Z → 2026-11-07T23:29:52Z
• Total runs analyzed: 592 agent-executed runs across 207 workflows
• Detection-enabled runs: 537 (90.7%) — gh-aw-detection is true or absent (the default)
• Regular runs: 55 (9.3%) — gh-aw-detection: false explicitly set
• Misconfigured workflows found: 0

Note

No misconfigured workflows were detected in this window. All 55 "regular" (detection-disabled) runs belong to one-off Smoke * test-harness workflows with a single run each in the window — none crossed the >3-runs-in-7-days threshold used to flag an actively-used workflow with detection turned off. No detection job showed a failing conclusion, and no workflow alternated between enabled/disabled settings within the window.

Comparison Chart

Detection-enabled runs ran at a higher success rate (89.9% vs 80.0%) but consumed roughly 3.3x the average tokens (441,710 vs 135,657) of regular runs — expected, since detection-enabled runs include heavier analytical/data-generation workflows (e.g. Daily Go Test Parallelizer, Ponytail Reviewer) while most gh-aw-detection: false runs are lightweight one-shot smoke tests.

Misconfigured Workflows

No misconfigured workflows detected in this window.

View All Run Metrics (top 30 of 207 workflows by run count)

Workflow	Detection Status	Runs	Success Rate	Avg Tokens
PR Sous Chef	enabled	130	100.0%	1,707
Test Quality Sentinel	enabled	37	97.3%	138,649
PR Code Quality Reviewer	enabled	34	94.1%	14,709
Matt Pocock Skills Reviewer	enabled	34	91.2%	82,377
Ponytail Reviewer	enabled	34	97.1%	602,826
Design Decision Gate 🏗️	enabled	32	90.6%	70,688
PR Specification Reviewer	enabled	31	93.5%	2,096
Daily Go Test Parallelizer	enabled	12	100.0%	9,162,851
Issue Monster	enabled	8	87.5%	75,182
Auditor	enabled	8	12.5%	0
Contribution Check	enabled	6	100.0%	27,165
Avenger	enabled	6	100.0%	405,562
Auto-Triage Issues	enabled	5	100.0%	397,596
Smoke Claude Dynamic Workflow	enabled	5	100.0%	1,158
[aw] Failure Investigator (6h)	enabled	5	80.0%	449
PR Triage Agent	enabled	4	100.0%	61,517
Code Scanning Fixer	enabled	4	75.0%	20,370
Daily Documentation Diagram	enabled	3	66.7%	2,672,950
Daily Cli Tools Tester	enabled	2	50.0%	2,876,426
Schema Consistency Checker	enabled	2	100.0%	4,365,372
Daily Credit Limit Test	enabled	2	100.0%	49,417
AI Moderator	enabled	2	100.0%	774
Visual Regression Checker	enabled	2	50.0%	7,931
Chaos PR Bundle Fuzzer	enabled	1	0.0%	1,466,202
Daily Model Inventory Checker	enabled	1	100.0%	21,589
Smoke Copilot SDK (and ~29 other Smoke * workflows)	disabled	1 each	varies	varies

The remaining ~180 workflows are single-run Smoke * CI/test-harness workflows (29 with gh-aw-detection: false, the rest with detection enabled) and a long tail of low-frequency scheduled/PR workflows.

View Historical Trend

39 daily data points on record (2026-10-09 → 2026-11-07, 26 of 30 recent days have data; gaps shown as breaks, not interpolated).

Recommendations

• No action needed for gh-aw-detection configuration this window — adoption (90.7% detection-enabled) and detection-job health are both healthy.
• Two active workflows outside the scope of this detection audit are worth separate follow-up: Auditor (8 runs, 12.5% success — recurring collect_failure_evidence step failures, see cluster failure-d317ca089a5bb99c) and Daily Documentation Diagram / Daily Cli Tools Tester (success rates 66.7%/50.0% with very high token usage). These are general reliability issues, not detection misconfigurations.
• Continue monitoring the Smoke * workflows with gh-aw-detection: false — they're intentionally detection-disabled smoke tests today, but if any of them become part of regular scheduled/active usage (>3 runs/week), re-evaluate whether detection should be re-enabled.

Coverage Note

Log collection hit the agentic-workflows MCP server's per-request API count limit (3000) twice while paging backward through the 24h window, requiring one follow-up continuation fetch. After that fetch, storage reached ~977 MB of the 1 GB budget with 947 run directories downloaded (592 agent-executed + 355 activation-only/non-execution runs). A small ~29-minute gap remains at the oldest edge of the window (2026-11-06T23:13:46Z–23:42:55Z) that was not fully re-fetched due to the storage budget. This represents ~98% coverage of the 24-hour window; findings above reflect this near-complete subset, not a guaranteed 100% census.

References:

• §42420748643 — Auditor, representative collect_failure_evidence failure
• §42382153785 — Matt Pocock Skills Reviewer, example detection-enabled run
• §42313775232 — PR Specification Reviewer, example activation-only (non-execution) run

Generated by 🔎 Detection Analysis Report · claude · agent · ~107 AIC · ⌖ ~1 AIC · ⊞ 4.5K · ◷

• expires on Nov 10, 2026, 3:36 PM UTC-08:00

[github-actions[bot]]

This discussion has been marked as outdated by Detection Analysis Report.

A newer discussion is available at Discussion #76005.