External-detector runs analysed: 555 · Detection-job error rate: 0.36% · Soft failures: 3 · Threat rate: 0% (of 144 runs with a verdict)

⚠️ Data is a partial-day sample, not a lower bound — collection was truncated:

verdict fetching stopped after 147/412 runs (budget exhausted)

Summary

gh-aw detection statistics - 2026-08-17 (UTC)

Repository: github/gh-aw
Window: 2026-08-17T00:00:00Z .. 2026-08-17T23:59:59Z
API requests: 1531, rate-limit pauses: 0
Data complete: NO - see Truncations

Totals

Metric	Count
Workflow runs in window	2623
Agentic runs (`*.lock.yml`)	1209
Runs with a `detection` job	780
... using the external detector	555
... using the built-in detector	75
... detector could not be determined	150
Agentic runs without a `detection` job	429

All rates below are over the external detector population (555 runs). A run is external when its detection job showed the Install threat-detect binary step, or when another run of the same workflow did that day — a job that was skipped or died during setup reports no steps and would otherwise be miscounted.

The rates below cover only the inspected part of the day. Collection was truncated, and the inspected runs are the earlier ones, so these percentages are a partial-day sample — not a lower bound.

Detection job outcomes

Outcome	Count	Rate
`success`	406	73.15%
`skipped`	143	25.77%
`cancelled`	4	0.72%
`failure`	2	0.36%

Error rate (failure/timed_out/action_required): 0.36%

Verdict availability

State	Count
`not_fetched`	408
`present`	144
`absent`	3

Green detection jobs that published no verdict: 3 (detection steps are continue-on-error, so a missing verdict artifact is the only reliable signal for these).

Detection results

Result	Count
Runs with a parsed verdict	144
Clean (no threat)	144
Any threat	0
`prompt_injection`	0
`secret_leak`	0
`malicious_patch`	0

Threat rate (of runs with a verdict): 0%

By workflow

Workflow	Runs	Failed	Cancelled	Skipped
PR Sous Chef	76	1	1	0
Deployment Incident Monitor	65	0	0	64
Issue Monster	47	0	0	26
Test Quality Sentinel	34	0	0	6
Matt Pocock Skills Reviewer	33	0	0	6
Ponytail Reviewer	33	0	0	7
Impeccable Skills Reviewer	32	0	0	7
PR Code Quality Reviewer	32	0	1	5
Design Decision Gate 🏗️	29	0	0	6
Avenger	24	1	0	7
PR Description Updater	20	0	1	7
Auto-Triage Issues	7	0	1	1
Daily Go Test Parallelizer	7	0	0	0
Contribution Check	6	0	0	0
Code Scanning Fixer	4	0	0	0
Deep Report	4	0	0	0
[aw] Failure Investigator (6h)	4	0	0	0
Daily Credit Limit Test	2	0	0	1
1. List all Go packages with their doc comments	1	0	0	0
Agent Job Health Monitor	1	0	0	0
Agent Performance Analyzer - Meta-Orchestrator	1	0	0	0
Agentic Workflow Audit Agent	1	0	0	0
Architecture Guardian	1	0	0	0
Archivx — Animated Workflow Visualizer	1	0	0	0
CI Optimization Coach	1	0	0	0

89 further workflows omitted; see stats.json.

Notable runs

Workflow	Run	Job conclusion	Verdict	Threats	Reported reason
Avenger	32042750467	`failure (Set up job)`	`not_fetched`	-	`-`
PR Sous Chef	32042666479	`failure (Set up job)`	`not_fetched`	-	`-`
Daily Safe Outputs Conformance Checker	31999265678	`success`	`absent`	-	`-`
Daily Sub-Agent Model Resolution Audit	31998521926	`success`	`absent`	-	`-`
arXiv Paper Researcher: GitHub Agentic Workflows	32011099290	`success`	`absent`	-	`-`

Truncations

The collector hit a budget or API limit; the numbers above are a lower bound.

verdict fetching stopped after 147/412 runs (budget exhausted)

Change since previous

No history yet; establishing the baseline. (The only prior cache entry was for this same date, 2026-08-17, and has been replaced with today's corrected figures.)

Watch list

PR Sous Chef — 1 failed, 0 without a verdict, out of 76 runs
Avenger — 1 failed, 0 without a verdict, out of 24 runs

Collected by: https://github.com/github/gh-aw-threat-detection/actions/runs/32180828286
Full data: the detection-stats-32180828286 artifact on that run.

Generated by Detection Stats Daily · auto · 22.5 AIC · ⌖ 3.21 AIC · ⊞ 10.6K · ◷