[detection-analysis] Detection Analysis Report — 2026-10-07

[github-actions[bot]]

Summary

• Window evaluated: logs covering 2026-10-07T14:35Z – 2026-10-07T23:19Z (~8h45m). This is the full span of pre-downloaded run logs available to this analysis; it is narrower than the nominal 24h window.
• Total runs analyzed: 311 (175 with full per-run telemetry, 136 with GitHub Actions run metadata only — see data-availability note below)
• Of the 175 classifiable runs: Detection-enabled runs: 175 (100%) · Regular (detection explicitly disabled) runs: 0 (0%)
• Misconfigured workflows found: 0

Note

No misconfigured gh-aw-detection workflows were detected in this window. Every classifiable run in the window had detection enabled (default or explicit true); no workflow explicitly disabled detection (gh-aw-detection: false) while running, so rules 1 and 3 found nothing to flag. Every detection job that executed completed successfully (166 success / 3 skipped, 0 failures), so rule 2 found nothing to flag either.

Comparison Chart

Because zero regular (detection-disabled) runs occurred in this window, the "Regular Runs" group is empty (0 runs, 0% success rate, 0 avg tokens) — this reflects the data, not a charting error. The repo does have workflows that set gh-aw-detection: false in source (59 .md files, mostly smoke-* and engine-conformance-* harnesses that intentionally skip detection for isolated engine testing), but none of them executed during the observed window.

Detection-enabled runs: 175 total, 141 succeeded (80.6%), 34 failed, averaging ~113,456 tokens/run.

Misconfigured Workflows

No misconfigured workflows detected in this window.

View All Run Metrics

Aggregate metrics (175 classifiable runs):

Metric	Regular	Detection
Run count	0	175
Success rate	—	80.6%
Failure count	0	34
Avg tokens/run	—	113,456.2

Per-workflow breakdown (top 15 by run count, all detection-enabled):

Workflow	Runs	Success %	Avg Tokens
[aw] Failure Investigator (6h)	16	100.0%	431
Issue Monster	15	100.0%	143,445
PR Sous Chef	13	100.0%	1,947
Test Quality Sentinel	12	91.7%	11,331
Design Decision Gate 🏗️	12	91.7%	1,564
Matt Pocock Skills Reviewer	11	90.9%	34,273
PR Code Quality Reviewer	11	90.9%	868,845
Impeccable Skills Reviewer	10	90.0%	32,120
Ponytail Reviewer	10	100.0%	114,617
Smoke Claude Dynamic Workflow	7	85.7%	1,207
Daily Go Test Parallelizer	5	0.0%	27,306
Smoke Gemini	2	0.0%	0
Avenger	2	0.0%	70,421
AI Moderator	2	100.0%	932
Auto-Triage Issues	2	100.0%	114,045

Plus 45 additional workflows with 1 run each in this window (full list available on request).

Data-availability note: 136 of the 311 total runs in the window had no aw_info.json/run_summary.json telemetry downloaded (GitHub Actions run metadata only), so their detection status could not be classified. Of these, 81 failed, 51 succeeded, 3 had no conclusion recorded, and 1 was a startup failure. The most frequent workflows in this bucket were Daily Trajectory Grader Implementer (18 runs), PR Sous Chef (18), AI Moderator (15), PR Code Quality Reviewer (12), Impeccable Skills Reviewer (11), Ponytail Reviewer (11), Test Quality Sentinel (11), and Matt Pocock Skills Reviewer (11). This is a log-collection gap (missing per-run telemetry artifacts), distinct from a detection misconfiguration — it is called out here for visibility but not counted toward the misconfigured-workflow findings above.

View Historical Trend

Trend computed from 7 daily snapshots within the last 30 days in cache-memory history (trending/detection-metrics/history.jsonl, 13 total entries since 2026-09-01, 6 older than the 30-day window excluded).

Recommendations

1. No action needed on gh-aw-detection configuration — all actively-running scheduled/triggered workflows in this window use detection in its default (enabled) state, and no detection job failures were observed.
2. Investigate the log-collection gap: 136 runs (44% of all runs in the window) lack per-run telemetry (aw_info.json/run_summary.json), including 81 outright failures. Workflows like daily-trajectory-grader-implementer, pr-sous-chef, ai-moderator, and the skills-reviewer family account for most of this gap and are worth checking for early-stage failures (e.g., setup/build steps before the agent job starts) that prevent telemetry generation — this masks visibility into whether detection is configured correctly for those runs.
3. Daily Go Test Parallelizer (0% success, 5/5 runs failed) and Avenger (0% success, 2/2 runs failed) stand out as reliability outliers among the classifiable detection-enabled runs; worth a follow-up investigation independent of detection configuration.

Generated by 🔎 Detection Analysis Report · claude · agent · 191.9 AIC · ⊞ 6.4K · ◷

• expires on Oct 10, 2026, 3:34 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 #67024.