[detection-analysis] Detection Analysis Report — 2026-08-09

[github-actions[bot]]

Summary

Analysis window: 2026-08-08 23:30 UTC → 2026-08-09 23:30 UTC (last 24 full hours)

Metric	Value
Total workflow runs	39
Detection-enabled runs	11 (28.2%)
Regular runs	28 (71.8%)
Detection success rate	90.9%
Regular success rate	25.0%
Misconfigured workflows	2

Warning

2 workflows were flagged as misconfigured this run: Q and ESLint Monster. Both run frequently (5 and 7 times respectively in the last 7 days) but have no gh-aw-detection feature configured at all. See the table below for details.

Note on the Regular-run success rate: of the 21 "failed" Regular runs, all 21 are action_required (blocked pending manual approval via the security gate for external/fork-triggered events) — not agent execution failures. Zero Regular runs had a true agent-logic failure. The one true failure in this window (PR Sous Chef, run 31338952148) was in the detection-enabled group and showed no detection-step-specific error signal (generic agent_logic_failure, ErrorCount=0).

Comparison Chart

Misconfigured Workflows

Workflow	Misconfiguration Type	Runs (7d)	Recommended Fix
Q	Rule 1: gh-aw-detection unset, >3 runs/7d (no features block at all in .github/workflows/q.md)	5	Add features:\n gh-aw-detection: true to the workflow frontmatter and recompile (gh aw compile)
ESLint Monster	Rule 1: gh-aw-detection unset, >3 runs/7d (no features block in .github/workflows/eslint-monster.md)	7	Add features:\n gh-aw-detection: true to the workflow frontmatter and recompile (gh aw compile)

No violations were found for Rule 2 (name-based opt-in expectation — all workflows named with audit/analyzer/report/detector/monitor/inspector already have detection enabled), Rule 3 (detection-step failures), or Rule 4 (alternating detection status within the window).

View All Run Metrics

Workflow	Engine	Detection	Runs (24h)	Outcomes
PR Sous Chef	pi	✅ true	7	6 success, 1 failure
Agentic Workflow Audit Agent	claude	✅ true	1	1 success
Daily Regulatory Report Generator	copilot	✅ true	1	1 success
Duplicate Code Detector	codex	✅ true	1	1 success
Daily AWF Spec Compiler Surfacing Review	pi	✅ true	1	1 success
AI Moderator	—	❌ false*	12	12 action_required (blocked, not failed)
Q	—	❌ false (misconfigured)	9	9 action_required (blocked, not failed)
Issue Monster	—	❌ false*	4	4 success
Avenger	—	❌ false*	2	2 success
ESLint Monster	pi	❌ false (misconfigured)	1	1 success

* AI Moderator, Issue Monster, and Avenger show detect=false in this window's run logs only because their agent job did not execute (approval gate or a job-level if: skip-condition) — their source frontmatter was verified via GitHub MCP to already have gh-aw-detection: true configured correctly. These are not misconfigurations.

Aggregate metrics:

• Regular: 28 runs, 25.0% success rate, 21 blocked (action_required), 0 true failures, avg tokens 715.4
• Detection-enabled: 11 runs, 90.9% success rate, 1 true failure, avg tokens 231013.2

View Historical Trend

Insufficient history to render a trend chart this run. The canonical trend cache (trending/detection/history.jsonl) now holds its first data point (today, 2026-08-09). A trend chart requires 2+ points and will be generated starting with tomorrow's run.

Data hygiene note: a differently-named cache file (trending/detection-metrics/history.jsonl) was found containing a single older record from 2026-08-08 with a different, incompatible schema (missing failure-count fields, and substantially different regular_runs/regular_avg_tokens values). It was not merged into the canonical history to avoid introducing inconsistent trend data. This naming/schema inconsistency should be reconciled in a future update to the trending logic.

Recommendations

1. Fix Q and ESLint Monster: add features: gh-aw-detection: true to both workflows' frontmatter and recompile, so future runs are properly classified and monitored for detection-step health.
2. Reconcile the trending cache: two divergent history files exist under /tmp/gh-aw/cache-memory/trending/ (detection/ vs detection-metrics/) with incompatible schemas — standardize on one path/schema so historical trending can accumulate reliably.
3. No action needed on AI Moderator, Issue Monster, or Avenger — their apparent detect=false status this window is a run-log artifact (job skipped/blocked), not a real misconfiguration; source frontmatter is already correct.
4. Monitor PR Sous Chef's single failure (run 31338952148) — no detection-specific error was found, but keep an eye out if failures recur, since it's the workflow with the most detection-enabled runs (7) in this window.
   (/content)

Generated by 🔎 Detection Analysis Report · agent · 247.5 AIC · ⌖ 49.9 AIC · ⊞ 7.4K · ◷

• expires on Aug 12, 2026, 3:36 PM UTC-08:00