[detection-analysis] Detection Analysis Report — 2026-11-04

[github-actions[bot]]

Summary

• Window evaluated: last 24 full hours ending at workflow start (UTC) — nominally 2026-11-03T23:15:16Z to 2026-11-04T23:12:34Z
• Coverage is partial: log collection hit the 1 GB storage budget after four paginated fetches. The analyzed subset actually spans 2026-11-04T04:04:10Z to 2026-11-04T23:12:34Z (~19.1 of the 24 hours). The oldest ~4h45m of the window (2026-11-03T23:15:16Z–2026-11-04T04:04:10Z) was not collected and is not reflected below.
• Total runs analyzed: 619 executed runs with usable aw_info.json/run_summary.json data, across 145 distinct workflows (an additional 208 run directories were non-execution/metadata-only — e.g. runs where the agent job never started — and were excluded from detection classification)
• Detection-enabled runs: 617 (99.7%)
• Regular runs (gh-aw-detection: false): 2 (0.3%)
• Misconfigured workflows found: 0

Note

No misconfigured workflows were detected in this window. The detection job showed no genuine failing conclusions (failure/cancelled/timed_out) anywhere in the analyzed data — the only non-success/skipped conclusions observed were empty strings on 3 runs, which corresponded to detection jobs still in_progress at snapshot time, not failures, so they were correctly excluded.

Comparison Chart

Metric	Regular Runs	Detection Runs
Run count	2	617
Success rate	100.0%	89.3%
Failure count	0	66
Avg tokens	6,846	257,056

The regular-run sample is too small (n=2) to be statistically meaningful; both workflows that disabled detection each ran only once in the analyzed window, so the 100% success rate is not a reliable comparison point. Detection-run average token usage is heavily skewed upward by a handful of very high-token workflows (see breakdown below), not by detection overhead itself.

Misconfigured Workflows

No misconfigured workflows detected in this window.

Checked and cleared:

• Rule 1 (gh-aw-detection: false with >3 runs): only 2 runs total had detection explicitly disabled — "Daily GitHub Docs SEO Optimizer" (§41775091564) and "Smoke Claude Routed" (§41927920550) — each with only 1 run, below the threshold.
• Rule 2 (detection job itself fails): zero runs had a detection job with a genuine failing conclusion.
• Rule 3 (alternates enabled/disabled within window): no workflow showed mixed detection settings across its runs.

View All Run Metrics

Top 25 workflows by run count (full breakdown covers all 145 workflows seen; most of the remainder had 1 run each):

Workflow	Total	Detection	Regular	Success %	Avg Tokens
PR Sous Chef	100	100	0	79.0	2,564
PR Finisher	58	58	0	91.4	12,138
Test Quality Sentinel	57	57	0	98.2	150,115
Matt Pocock Skills Reviewer	54	54	0	92.6	73,908
PR Code Quality Reviewer	54	54	0	98.1	17,063
Ponytail Reviewer	53	53	0	98.1	570,283
Design Decision Gate 🏗️	52	52	0	98.1	70,390
Avenger	13	13	0	84.6	644,129
Daily Go Test Parallelizer	10	10	0	100.0	5,655,430
AI Moderator	9	9	0	88.9	1,038
Issue Monster	7	7	0	100.0	70,594
Auto-Triage Issues	6	6	0	100.0	269,064
Contribution Check	5	5	0	100.0	28,466
[aw] Failure Investigator (6h)	4	4	0	75.0	311
Code Scanning Fixer	3	3	0	66.7	20,708
PR Triage Agent	3	3	0	100.0	59,367
Daily Elixir Credo Snippet Audit	2	2	0	100.0	13,250
Visual Regression Checker	2	2	0	100.0	7,712
AW Essential Issue Clustering	1	1	0	100.0	14,891
Actions and AW Linter Factory	1	1	0	100.0	95,390
Agent Job Health Monitor	1	1	0	0.0	0
Agent Performance Analyzer - Meta-Orchestrator	1	1	0	100.0	20,730
Agent Persona Explorer	1	1	0	100.0	29,234
Agentic Workflow AIC Usage Optimizer	1	1	0	100.0	14,220
Agentic Workflow Audit Agent	1	1	0	100.0	2,582,187

Note: "Daily Go Test Parallelizer" and "Agentic Workflow Audit Agent" show unusually high average token usage (5.6M and 2.6M respectively) — worth a separate look if token budgets are a concern, though neither trips a detection-misconfiguration rule.

View Historical Trend

Today's metrics were appended to the cache-memory trending history (36 data points now on record, spanning a 30-day window with some gap days). Detection success rate and run counts have been broadly stable; no new anomalies surfaced in the trend.

Recommendations

1. No misconfiguration fixes needed this cycle — detection is enabled by default across 99.7% of runs, and no detection job failures or inconsistent per-workflow settings were observed.
2. Re-run coverage for the missing ~4h45m window (2026-11-03T23:15:16Z–2026-11-04T04:04:10Z) in a follow-up pass with a larger storage budget or narrower per-call count/max_storage, since this report's totals exclude that slice.
3. Consider investigating token-usage outliers ("Daily Go Test Parallelizer", "Agentic Workflow Audit Agent", "Ponytail Reviewer") separately — their multi-hundred-thousand-to-multi-million token averages are the main driver of the detection-run average-token gap shown in the comparison chart, not detection overhead itself.
4. The two workflows with gh-aw-detection: false ("Daily GitHub Docs SEO Optimizer", "Smoke Claude Routed") are each below the 3-run flagging threshold this cycle; monitor them in future windows in case their run frequency increases.

References:

• §41925400471 — Avenger (agent-logic failure cluster, part of prior failure investigation)
• §41775091564 — Daily GitHub Docs SEO Optimizer (one of two gh-aw-detection: false runs this window)
• §41927920550 — Smoke Claude Routed (second gh-aw-detection: false run this window)

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

• expires on Nov 7, 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 #75254.