[detection-analysis] Detection Analysis Report — 2026-10-28

[github-actions[bot]]

Summary

• Window evaluated: last 24 full hours (UTC), 2026-10-27T23:16:39Z → 2026-10-28T23:16Z
• Total runs analyzed: 457 (with full run detail) out of 588 total run records in the window
• Detection-enabled runs: 440 (96.3%)
• Regular runs (gh-aw-detection: false): 17 (3.7%)
• Misconfigured workflows found: 0 confirmed

Note

No confirmed detection misconfigurations in this window. 131 run records (obsolete-PR / non-execution runs) lacked full aw_info.json/run_summary.json detail and were excluded from detection classification — coverage of those is incomplete by design (metadata-only).

Comparison Chart

Metric	Regular runs	Detection runs
Run count	17	440
Success rate	76.5%	75.5%
Failure count	4	108
Avg. token usage	84,094	140,986

Misconfigured Workflows

No misconfigured workflows detected in this window.

All 17 "regular" (detection-disabled) runs were single, isolated runs of smoke-test workflows (Smoke Claude, Smoke Codex, Smoke Copilot*, Smoke Pi*, Smoke OTEL, Smoke Agy (Experimental), Agent Container Smoke Test) plus one occurrence of Daily GitHub Docs SEO Optimizer. None showed more than 1 run in this 24h window, so rule 1 (gh-aw-detection: false with >3 runs/7 days) could not be confirmed from this data — a single 24h snapshot cannot establish a 7-day run-count threshold, and projecting from one observed run would be speculative. No workflow's detection job itself failed (rule 2), and no workflow alternated between enabled/disabled within the window (rule 3).

Engine/workflow mix

No detection-job failures were observed across 439 runs that had a detection job present in job_details (424 success, 15 skipped for detection-disabled runs).

View All Run Metrics

Top workflows by run count in the window (176 distinct workflows total; showing top 20 plus all 17 regular/detection-disabled workflows):

Workflow	Runs	Mode	Success%	Avg Tokens	Engine(s)
PR Sous Chef	114	Detection	40.4%	2,235	pi
Issue Monster	47	Detection	89.4%	68,628	pi
Avenger	24	Detection	95.8%	275,247	copilot
Test Quality Sentinel	13	Detection	100.0%	13,527	copilot
Daily Go Test Parallelizer	12	Detection	91.7%	1,326,336	codex
Matt Pocock Skills Reviewer	12	Detection	100.0%	47,199	copilot
PR Code Quality Reviewer	12	Detection	100.0%	16,360	copilot
Ponytail Reviewer	12	Detection	100.0%	160,915	codex
AI Moderator	11	Detection	100.0%	881	pi
Design Decision Gate 🏗️	11	Detection	100.0%	1,485	pi
Contribution Check	6	Detection	100.0%	16,201	copilot
Auto-Triage Issues	5	Detection	100.0%	136,724	codex
[aw] Failure Investigator (6h)	5	Detection	100.0%	415	claude
PR Triage Agent	4	Detection	100.0%	75,697	copilot
Code Scanning Fixer	4	Detection	100.0%	10,995	copilot
Visual Regression Checker	4	Detection	0.0%	18,507	copilot
Daily Credit Limit Test	2	Detection	100.0%	46,408	pi
Daily GitHub Docs SEO Optimizer	1	Regular	100.0%	10,482	copilot
Agent Container Smoke Test	1	Regular	100.0%	130,072	pi
Smoke Claude	1	Regular	100.0%	14,621	claude
Smoke Codex	1	Regular	100.0%	437,494	codex
Smoke Copilot Auto	1	Regular	100.0%	4,200	copilot
Smoke Pi	1	Regular	0.0%	23,899	pi
Smoke Agy (Experimental)	1	Regular	0.0%	0	agy
Smoke Claude Routed	1	Regular	100.0%	1,831	claude
Smoke Codex Routed	1	Regular	100.0%	52,238	codex
Smoke Copilot	1	Regular	100.0%	4,028	copilot
Smoke Copilot - AOAI (apikey)	1	Regular	0.0%	87,301	copilot
Smoke Copilot - AOAI (Entra)	1	Regular	0.0%	0	copilot
Smoke Copilot MAI	1	Regular	100.0%	63,073	copilot
Smoke Copilot SDK Routed	1	Regular	100.0%	4,191	copilot
Smoke Copilot Small	1	Regular	100.0%	2,290	copilot
Smoke OTEL	1	Regular	100.0%	532,764	codex
Smoke Pi Routed	1	Regular	100.0%	61,105	pi

View Historical Trend

30 days of history now available (18 of 30 days have data; gaps are left as breaks, not interpolated).

Recommendations

• No corrective action required — no misconfigured workflows were confirmed in this window.
• The lower success rate and higher token usage on detection-enabled runs is driven mainly by PR Sous Chef (114 runs, 40.4% success) and general failure volume (108 of 440 detection-enabled runs failing) rather than detection itself; worth a separate look at PR Sous Chef's failure pattern if not already tracked.
• Confirming the 7-day gh-aw-detection: false + >3-runs rule needs a 7-day log fetch, which is out of scope for a 24h window — consider running this check on a longer cadence (e.g., weekly) if that misconfiguration pattern is a priority.

References:

• §40667284360 — PR Sous Chef failure example (safe_outputs/"Process Safe Outputs" step)
• §40765587560 — Daily Caveman Optimizer failure example
• §40687536087 — Smoke Agy (Experimental) run (regular/detection-disabled)

Generated by 🔎 Detection Analysis Report · claude · agent · ~109 AIC · ⌖ ~1 AIC · ⊞ 4.4K · ◷

• expires on Oct 31, 2026, 3:33 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 #73057.