[detection-analysis] Detection Analysis Report — 2026-08-18

[github-actions[bot]]

Summary

• Window evaluated: last 24 full hours (UTC), 2026-08-17T23:26:08Z to 2026-08-18T23:26:08Z
• Total runs analyzed: 298 (from 298 of 702 discovered run directories that contained logs)
• Detection-enabled runs: 259 (86.9%)
• Regular runs: 39 (13.1%)
• Misconfigured workflows found: 2

Warning

2 workflows with meaningful run volume have gh-aw-detection disabled/absent despite defining safe-outputs.

Comparison Chart

Misconfigured Workflows

Workflow	Misconfiguration Type	Run Count	Recommended Fix
Squad	detection disabled on high-volume workflow (>3 runs, rule 1)	8	Add features:\n gh-aw-detection: true to .github/workflows/squad.md — it defines safe-outputs and runs repeatedly via workflow_dispatch fan-out. Example run: §32133947135
Squad Implement Worker	detection disabled on high-volume workflow (>3 runs, rule 1)	11	Add features:\n gh-aw-detection: true to .github/workflows/squad-implement-worker.md — it defines safe-outputs and is invoked frequently as the Squad orchestrator's sub-agent. Example run: §32160700508

No matches were found for keyword-based (rule 2), detection-step-failure (rule 3), or alternating-config (rule 4) misconfiguration patterns in this window — every keyword-matching workflow (audit/analyzer/report/detector/monitor/inspector) already had gh-aw-detection: true, every run's threat-detection artifact was present regardless of overall job outcome, and no workflow name showed mixed detection settings across its runs.

Daily Agentic Workflow AIC Usage Audit did not run in this window; it remains excluded from rule 2 per its documented upstream-mirrored opt-out and was not evaluated further.

View All Run Metrics

Aggregate:

• Regular runs: 39 — success rate 92.3%, avg tokens 24,671
• Detection runs: 259 — success rate 83.4%, avg tokens 54,772

Note: the detection-runs success rate is pulled down mainly by PR Sous Chef (40 runs, 10% success) — a known chronic-failure workflow (see agent-job-health.md), unrelated to the detection feature itself.

Full per-workflow breakdown (80 distinct workflows):

Workflow	Detection	Runs	Success %	Avg Tokens	Engine(s)
PR Sous Chef	✅ true	40	10.0%	50,941	pi
Test Quality Sentinel	✅ true	24	100.0%	7,254	copilot
Ponytail Reviewer	✅ true	22	100.0%	9,630	copilot
Impeccable Skills Reviewer	✅ true	21	100.0%	7,787	copilot
Matt Pocock Skills Reviewer	✅ true	21	100.0%	14,931	copilot
Design Decision Gate 🏗️	✅ true	21	100.0%	13,239	claude
PR Code Quality Reviewer	✅ true	21	100.0%	41,944	pi
Issue Monster	✅ true	12	100.0%	20,773	pi
Squad Implement Worker	❌ false/absent	11	100.0%	27,041	copilot
Squad	❌ false/absent	8	100.0%	32,122	copilot
AI Moderator	✅ true	6	100.0%	141,498	codex
Daily Go Test Parallelizer	✅ true	6	100.0%	339,280	copilot
PR Description Updater	✅ true	5	100.0%	9,438	copilot
Avenger	✅ true	4	100.0%	10,315	claude
Auto-Triage Issues	✅ true	4	100.0%	24,273	pi
Contribution Check	✅ true	3	100.0%	25,797	copilot
Deep Report	✅ true	2	100.0%	55,269	claude
Code Scanning Fixer	✅ true	2	100.0%	9,867	copilot
PR Triage Agent	❌ false/absent	2	100.0%	23,510	copilot
[aw] Failure Investigator (6h)	✅ true	2	100.0%	25,144	claude
Scout	❌ false/absent	2	0.0%	0	claude
Daily Spec Coverage Review	✅ true	1	100.0%	8,096	copilot
Agentic Workflow Audit Agent	✅ true	1	0.0%	114,325	claude
Daily Copilot PR Merged Report	✅ true	1	100.0%	10,522	copilot
The Daily Repository Chronicle	✅ true	1	100.0%	27,446	copilot
Daily Testify Uber Super Expert	✅ true	1	100.0%	11,630	copilot
Slide Deck Maintainer	❌ false/absent	1	100.0%	33,226	copilot
Squad Game Planner	❌ false/absent	1	100.0%	29,019	copilot
Daily Safe Output Tool Optimizer	✅ true	1	100.0%	62,898	claude
Architecture Guardian	✅ true	1	0.0%	0	copilot
Daily Project Performance Summary Generator (Using MCP Scripts)	✅ true	1	100.0%	27,554	copilot
Dead Code Removal Agent	❌ false/absent	1	100.0%	60,739	copilot
Package Specification Librarian	❌ false/absent	1	100.0%	11,405	copilot
GitHub MCP Structural Analysis	✅ true	1	100.0%	29,144	claude
Daily Go Test Stubs — Aider	✅ true	1	100.0%	5,182	aider
Daily Safe Output Integrator	✅ true	1	100.0%	10,619	copilot
Agentic Workflow AIC Usage Optimizer	❌ false/absent	1	100.0%	23,562	copilot
Daily AWF Spec Compiler Surfacing Review	✅ true	1	100.0%	0	pi
GEO Optimizer Daily Audit	✅ true	1	100.0%	11,872	copilot
Daily SPDD Spec Planner	✅ true	1	100.0%	14,065	copilot
Duplicate Code Detector	✅ true	1	100.0%	3,903,758	codex
Daily Cache Strategy Analyzer	✅ true	1	100.0%	1,811,750	codex
Daily Code Debt Cleanup — Aider	✅ true	1	100.0%	4,971	aider
Daily Caveman Optimizer	✅ true	1	100.0%	12,928	claude
Daily Security Observability Report	✅ true	1	100.0%	32,673	copilot
Daily CLI Performance Agent	✅ true	1	100.0%	27,141	copilot
Daily Ambient Context Optimizer	✅ true	1	100.0%	27,879	copilot
Release	❌ false/absent	1	100.0%	12,877	copilot
Agent Persona Explorer	✅ true	1	100.0%	36,998	pi
Daily Formal Spec Verifier	✅ true	1	100.0%	34,372	copilot
Lockfile Statistics Analysis Agent	❌ false/absent	1	100.0%	26,339	claude
Smoke Copilot	❌ false/absent	1	100.0%	15,084	copilot
ESLint Monster	✅ true	1	100.0%	21,187	pi
Issue Triage Agent	❌ false/absent	1	100.0%	9,604	copilot
Daily Issues Report Generator	✅ true	1	100.0%	21,641	copilot
Repository Quality Improvement Agent	❌ false/absent	1	100.0%	17,521	copilot
Claude Code User Documentation Review	✅ true	1	100.0%	48,462	claude
Daily AIC Consumption Report (Sentry + Grafana OTel)	✅ true	1	100.0%	14,460	goose
PureLock	❌ false/absent	1	100.0%	20,294	copilot
Delight	✅ true	1	100.0%	24,877	copilot
Daily Reliability Review	✅ true	1	100.0%	10,100	opencode
Daily Fact	✅ true	1	0.0%	0	codex
Daily Secrets Analysis Agent	❌ false/absent	1	100.0%	11,542	copilot
Workflow Normalizer	❌ false/absent	1	100.0%	15,521	copilot
Daily Regulatory Report Generator	✅ true	1	100.0%	18,766	copilot
Linter Miner	❌ false/absent	1	0.0%	29,420	copilot
Daily Malicious Code Scan Agent	✅ true	1	0.0%	0	copilot
Daily Credit Limit Test	✅ true	1	100.0%	4,805	copilot
Daily File Diet	✅ true	1	100.0%	24,891	copilot
Daily BYOK Ollama Test	✅ true	1	0.0%	0	copilot
CLI Consistency Checker	✅ true	1	0.0%	0	copilot
UK AI Operational Resilience	❌ false/absent	1	100.0%	25,496	copilot
Copilot PR Prompt Pattern Analysis	✅ true	1	100.0%	16,472	copilot
CI Optimization Coach	✅ true	1	0.0%	0	copilot
Daily Code Metrics and Trend Tracking Agent	✅ true	1	100.0%	35,981	crush
Daily Graft Intelligence	✅ true	1	100.0%	7,811	copilot
Agent Performance Analyzer - Meta-Orchestrator	✅ true	1	100.0%	28,676	copilot
Smoke Copilot Sub Agents	❌ false/absent	1	100.0%	19,081	copilot
Daily Agent of the Day Blog Writer	✅ true	1	100.0%	22,624	copilot
Daily Team Evolution Insights	✅ true	1	100.0%	22,377	goose

View Historical Trend

No trend chart this run — this is the first recorded entry in the cache-memory trending history (/tmp/gh-aw/cache-memory/trending/detection-metrics/history.jsonl). A 30-day trend chart will be generated automatically once 7+ days of history have accumulated.

Recommendations

1. Enable gh-aw-detection: true on squad.md and squad-implement-worker.md — both produce safe-outputs and run frequently; leaving detection off skips prompt-injection scanning on their output.
2. Audit any future new workflow whose name suggests it inspects other workflows/agents (audit, analyzer, report, detector, monitor, inspector) — none violated this rule in the current window, but keep the check running as new workflows are added.
3. Separately from detection config, PR Sous Chef is failing in 90% of its runs (40 total) — this is a job-health issue, not a detection misconfiguration, but it's worth tracking since it dominates the "detection" cohort's token and failure totals.

References:

• §32160700508 — Squad Implement Worker (detection disabled)
• §32133947135 — Squad (detection disabled)
• §32186551970 — Agentic Workflow Audit Agent (failure, detection enabled — included for context, not flagged)

Generated by 🔎 Detection Analysis Report · agent · 149.5 AIC · ⌖ 13 AIC · ⊞ 7.8K · ◷

• expires on Aug 21, 2026, 3:38 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 #54091.