[detection-analysis] Detection Analysis Report — 2026-08-12

[github-actions[bot]]

Summary

• Window evaluated: last 24 full hours (UTC), 2026-08-11T23:33:00Z to 2026-08-12T23:33:00Z (approx., ending at this report's start)
• Total runs analyzed: 29 concluded (+2 still in progress at analysis time — this report's own run and Agent Job Health Monitor)
• Detection-enabled runs: 12 (41%)
• Regular runs: 17 (59%)
• Misconfigured workflows found: 0

Note

No misconfigured workflows detected in this window. See "Misconfigured Workflows" below for a pattern that looked suspicious at first glance but checked out as a telemetry gap, not a real misconfiguration.

Comparison Chart

Regular runs succeeded 4/17 (23.5%) vs. detection runs 10/12 (83.3%). Detection-enabled runs averaged 97,426 tokens; regular runs averaged 0 tokens.

Misconfigured Workflows

No misconfigured workflows detected in this window.

One pattern was investigated and ruled out: AI Moderator (7 runs), Issue Monster (3 runs), and Q (4 runs) each appear in the "Regular" bucket in this window even though their source configs (ai-moderator.md, issue-monster.md, q.md) all declare gh-aw-detection: true, and their compiled lock files confirm GH_AW_INFO_FEATURES: '{"gh-aw-detection":true}'. This looked like rule-4 alternation at first. On inspection, every one of those "Regular" runs is either action_required (blocked on a GitHub Actions approval gate before the agent job ever started, so aw_info.json was never written) or a 0-turn early-exit "success" (Issue Monster's skip path). None of them reflect an actual toggle of the detection feature — it's a telemetry gap where incomplete/gated runs don't emit detection metadata, not a config problem. No fix needed on the workflow side; flagging this as a possible observability improvement (see Recommendations).

Avenger and PR Description Updater/Q/AI Moderator's gated runs make up the rest of the Regular bucket; Avenger genuinely has no gh-aw-detection feature configured, but its name doesn't match the audit/analyzer/report/detector/monitor/inspector keyword list, so rule 2 doesn't apply.

View All Run Metrics

Workflow	Detection	Status	Tokens	Run
PR Code Quality Reviewer	✅ true	success	0	§31643652223
AI Moderator	✅ true	success	112,567	§31644275419
Issue Monster	✅ true	failure	19,605	§31644531765
PR Sous Chef	✅ true	success	54,342	§31644728185
Duplicate Code Detector	✅ true	success	832,269	§31645254067
ESLint Monster	✅ true	success	19,676	§31645591353
PR Sous Chef	✅ true	success	14,542	§31646687778
Daily AWF Spec Compiler Surfacing Review	✅ true	success	0	§31646760640
Daily BYOK Ollama Test	✅ true	failure	0	§31647146503
PR Sous Chef	✅ true	success	53,708	§31648061995
Daily Reliability Review	✅ true	success	9,593	§31649587636
PR Sous Chef	✅ true	success	52,807	§31649732812
Agent Job Health Monitor	✅ true	in progress	—	§31650574304
PR Description Updater	⬜ absent	action_required	0	§31643945397
AI Moderator ×6	⬜ absent	action_required	0	§31644890331 +5 more
Q ×4	⬜ absent	action_required	0	§31646222744 +3 more
Issue Monster ×3	⬜ absent	success (0-turn skip)	0	§31646587284 +2 more
Avenger	⬜ absent (no feature configured)	success	0	§31647813057

View Historical Trend

Not enough history yet — this run added the first entry to /tmp/gh-aw/cache-memory/trending/detection-metrics/history.jsonl. A trend chart will be generated automatically once 7+ daily entries have accumulated.

Recommendations

• Investigate the two detection-enabled failures — Issue Monster (§31644531765) and Daily BYOK Ollama Test (§31647146503) — each failed with 1 error. Neither failure traced to a detection-specific step in the available logs, but both are worth a closer look.
• Consider writing a minimal aw_info.json (or equivalent detection-feature marker) earlier in the job — before the approval-gate/early-exit paths — so action_required and 0-turn skip runs don't silently disappear from detection telemetry. This would remove the false "alternation" signal for AI Moderator, Issue Monster, and Q going forward.
• No workflow-level gh-aw-detection config changes needed this cycle.

References:

• §31644531765 — Issue Monster failure
• §31647146503 — Daily BYOK Ollama Test failure
• §31645254067 — highest token usage this window (Duplicate Code Detector, 832,269 tokens)

Generated by 🔎 Detection Analysis Report · agent · 161.9 AIC · ⌖ 44.3 AIC · ⊞ 7.4K · ◷

• expires on Aug 15, 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 #52595.