[detection-analysis] Detection Analysis Report — 2026-10-18

[github-actions[bot]]

Summary

• Window evaluated: last 24 full hours (UTC), approximately 2026-10-17T23:19:00Z to 2026-10-18T23:19:00Z
• Total run records downloaded: 636 (494 fully classified with aw_info.json + run_summary.json; 142 metadata-only, detection status unknown and excluded from the stats below)
• Detection-enabled runs: 473 (95.7% of classified runs)
• Regular runs (gh-aw-detection: false): 21 (4.3% of classified runs)
• Misconfigured workflows found: 1

Warning

1 workflow was found with gh-aw-detection explicitly disabled across all of its runs in this window.

Comparison Chart

Metric	Regular runs	Detection runs
Run count	21	473
Success rate	76.2%	92.6%
Failure count	5	35
Avg. token usage	347,373	125,432

Misconfigured Workflows

Workflow	Misconfiguration type	Run count	Recommended fix
Smoke Copilot SDK Routed	gh-aw-detection: false on every run (>3 runs in window)	4	This is a private: true smoke/routing-test workflow that sets features.gh-aw-detection: false in its frontmatter — likely intentional to reduce noise on a test workflow. Confirm the opt-out is deliberate; if not, remove the override so detection defaults back on. Example run: §39147798082

No detection-job step failures (rule 2) or inconsistent per-run detection settings (rule 3) were observed in this window.

View All Run Metrics

146 distinct workflows were observed. Top 20 by run count:

Workflow	Detection status	Runs	Success rate	Avg. tokens
PR Sous Chef	detection	98	99.0%	2,960
Issue Monster	detection	46	93.5%	129,704
Matt Pocock Skills Reviewer	detection	31	100.0%	43,616
PR Code Quality Reviewer	detection	27	100.0%	141,150
Ponytail Reviewer	detection	27	100.0%	138,088
Test Quality Sentinel	detection	27	100.0%	10,364
Design Decision Gate 🏗️	detection	26	100.0%	1,506
Daily Trajectory Grader Implementer	detection	21	95.2%	6,684
Daily Go Test Parallelizer	detection	12	91.7%	1,109,003
AI Moderator	detection	6	100.0%	835
Contribution Check	detection	6	100.0%	19,642
Smoke Copilot Routed	detection	6	66.7%	92,376
[aw] Failure Investigator (6h)	detection	6	66.7%	273
Auto-Triage Issues	detection	5	100.0%	156,954
Avenger	detection	4	75.0%	560,235
Code Scanning Fixer	detection	4	100.0%	10,841
PR Triage Agent	detection	4	100.0%	54,685
Smoke Copilot SDK Routed	regular	4	100.0%	2,577
Smoke Copilot Routed Claude	regular	3	66.7%	6,242
Smoke Copilot Routed Steering	regular	3	66.7%	2,366,495

Remaining ~126 workflows had 1-2 runs each in this window.

View Historical Trend

Based on 8 trending snapshots within the last 30 days (/tmp/gh-aw/cache-memory/trending/detection-metrics/history.jsonl, 20 entries total since 2026-08-25).

Recommendations

1. Confirm the Smoke Copilot SDK Routed opt-out is intentional. It's a private: true smoke-test workflow, so the disabled detection feature may be deliberate — but it's worth a one-line comment in the frontmatter documenting why, so future audits don't keep re-flagging it.
2. No detection-job failures observed in this window — the feature itself is running cleanly wherever it's enabled.
3. Coverage gap: 142 of 636 downloaded run records (22%) were metadata-only (no aw_info.json/run_summary.json), so their detection status is unknown and they were excluded from all stats above. This is a pre-existing collection/retention characteristic of older or smaller runs, not a workflow misconfiguration.

References:

• §39147798082 — example Smoke Copilot SDK Routed run with detection disabled
• §39120173514 — Issue Monster, representative failure cluster run
• §39154584813 — Daily Code Metrics and Trend Tracking Agent, representative failure cluster run

Generated by 🔎 Detection Analysis Report · claude · agent · 111.4 AIC · ⌖ 0.863 AIC · ⊞ 6.6K · ◷

• expires on Oct 21, 2026, 3:32 PM UTC-08:00