[detection-analysis] Detection Analysis Report — Partial 24h Window (2026-10-20)

[github-actions[bot]]

Summary

• Window requested: last 24 full hours (UTC) ending at workflow start (~2026-10-20T23:21Z)
• Actual coverage: only ~4.9 hours (2026-10-20T18:26:59Z to 2026-10-20T23:18:13Z) — log collection hit its count limit (3000) before backfilling the remaining ~19 hours of the window. This is a partial-window analysis, not a full 24h analysis.
• Total runs analyzed: 135 with full detail (121 detection-enabled, 14 regular) + 28 metadata-only runs (detection status unknown — only the GitHub run record was downloaded, no aw_info.json/run_summary.json)
• Detection-enabled runs: 121 (89.6% of classified runs)
• Regular runs (gh-aw-detection: false): 14 (10.4% of classified runs)
• Misconfigured workflows found: 1

Warning

Coverage gap: roughly 19 of the requested 24 hours (2026-10-19T23:21Z–2026-10-20T18:26Z) were never downloaded due to a collection count limit, not an API rate limit (13,242/15,000 GitHub API calls remained at the end of collection). Findings below reflect only the ~4.9h window that was actually downloaded. Misconfiguration rule 1 ("gh-aw-detection: false on a workflow with >3 runs") is specified against a 7-day window in the task definition but was evaluated here against the available ~4.9h sample only — treat the flagged workflow as a lead to verify over the full 7 days, not a confirmed 7-day finding.

Comparison Chart

Regular runs (14) had a 21.4% success rate and averaged 7,507 tokens; detection-enabled runs (121) had an 86.0% success rate and averaged 173,121 tokens. The regular-run bucket is entirely smoke-test workflows (see table below), so the gap partly reflects sample composition (smoke tests vs. production workflows), not detection itself.

Misconfigured Workflows

Workflow	Misconfiguration Type	Run Count	Recommended Fix
Smoke Copilot	gh-aw-detection: false explicitly set, >3 runs observed in the ~4.9h sample	10	Confirm this is intentional for a smoke-test workflow (detection may not be meaningful for synthetic test runs). If not intentional, remove gh-aw-detection: false from frontmatter and recompile. Example run: §39475833194

No detection-job-level failures (rule 2) or alternating detection configuration within a single workflow (rule 3) were found in the observed sample: of 121 detection-enabled runs, 117 had a detection job in job_details (110 succeeded, 7 skipped, 0 failed); the remaining 4 had no detection job recorded.

View All Run Metrics

Per-workflow breakdown for the ~4.9h observed window (40 workflows, 135 classified runs):

Workflow	Runs	Detection Status	Success Rate	Avg Tokens
PR Sous Chef	20	Enabled	100.0%	1,168
Ponytail Reviewer	11	Enabled	81.8%	116,198
Design Decision Gate 🏗️	11	Enabled	81.8%	7,067
Smoke Copilot	10	Disabled	20.0%	10,153
Issue Monster	10	Enabled	90.0%	77,838
Matt Pocock Skills Reviewer	10	Enabled	80.0%	46,283
PR Code Quality Reviewer	10	Enabled	80.0%	14,598
Test Quality Sentinel	10	Enabled	80.0%	10,351
Auto-Triage Issues	4	Enabled	100.0%	110,330
Avenger	4	Enabled	100.0%	858,990
Daily Go Test Parallelizer	3	Enabled	100.0%	1,759,347
Smoke Claude Dynamic Workflow	3	Enabled	100.0%	1,139
Smoke Copilot Sub Agents	2	Disabled	0.0%	0
PR Triage Agent	1	Enabled	100.0%	62,400
Copilot Agent PR Analysis	1	Enabled	100.0%	24,706
Daily Cache Strategy Analyzer	1	Enabled	100.0%	5,301,389
Daily Code Metrics and Trend Tracking Agent	1	Enabled	100.0%	16,894
Daily Safe Output Integrator	1	Enabled	100.0%	7,849
Daily Code Debt Cleanup — Aider	1	Enabled	100.0%	64,898
Code Scanning Fixer	1	Enabled	100.0%	10,834
Daily Spec Coverage Review	1	Enabled	0.0%	5,464
Daily Graft Intelligence	1	Enabled	100.0%	6,306
[aw] Failure Investigator (6h)	1	Enabled	100.0%	379
Daily Ambient Context Optimizer	1	Enabled	100.0%	18,649
Lockfile Statistics Analysis Agent	1	Enabled	100.0%	19,577
Daily Work Queue Formal Verification	1	Enabled	0.0%	0
Copilot PR Prompt Pattern Analysis	1	Enabled	100.0%	14,477
Daily Caveman Optimizer	1	Enabled	100.0%	6,393
Daily Project Performance Summary Generator (Using MCP Scripts)	1	Enabled	100.0%	239,644
Contribution Check	1	Enabled	100.0%	19,765
Daily Safe Output Tool Optimizer	1	Enabled	100.0%	33,559
Smoke Copilot SDK Routed Claude	1	Disabled	0.0%	1,136
Agentic Workflow Audit Agent	1	Enabled	100.0%	2,365,009
Duplicate Code Detector	1	Enabled	100.0%	687,895
Daily AWF Spec Compiler Surfacing Review	1	Enabled	0.0%	0
Daily BYOK Ollama Test	1	Enabled	0.0%	0
Smoke Copilot SDK Routed	1	Disabled	100.0%	2,426
Daily Reliability Review	1	Enabled	100.0%	12,800
Detection Analysis Report	1	Enabled	0.0%	0
Agent Job Health Monitor	1	Enabled	0.0%	0

28 additional runs in the window had only a GitHub run record downloaded (no aw_info.json/run_summary.json), so their detection status is unknown and they are excluded from the classification above (consistent with the "metadata-only groups remain unknown" rule).

View Historical Trend

10 of the last 30 days have recorded data points. Caveat: today's point reflects the same partial ~4.9h window described above, not a full day, so the sharp drop in both run counts on the latest date is a coverage artifact, not a real traffic decline — exclude it when judging day-over-day trend.

Recommendations

1. Verify whether gh-aw-detection: false on Smoke Copilot is intentional (it is a synthetic smoke-test workflow with a 20% success rate, which may be by design for exercising failure paths). If unintentional, remove the override and recompile.
2. Re-run detection analysis once full 24h log collection succeeds (raise the collection count/storage limit or page through the remaining ~19h) before drawing conclusions about the full day's detection coverage or success-rate comparison.
3. No detection-job failures were found in this sample — no remediation needed for rule 2 at this time.

References:

• §39475833194 — Smoke Copilot (flagged misconfiguration)
• §39507452713 — Daily AWF Spec Compiler Surfacing Review (failure cluster example)
• §39499464895 — Smoke Copilot SDK Routed Claude (failure cluster example)

Generated by 🔎 Detection Analysis Report · claude · agent · 166 AIC · ⌖ 0.86 AIC · ⊞ 6.6K · ◷

• expires on Oct 23, 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 #71243.