[detection-analysis] Detection Analysis Report — 2026-08-24

[github-actions[bot]]

Summary

• Window evaluated: last 24 full hours (UTC), 2026-08-23T23:28Z → 2026-08-24T23:28Z
• Total runs analyzed: 267 (across 67 distinct workflows)
• Detection-enabled runs: 263 (98.5%)
• Regular runs (gh-aw-detection: false): 4 (1.5%)
• Misconfigured workflows found: 0

Note

No misconfigured workflows were detected in this window. All three misconfiguration rules (persistent opt-out on high-volume workflows, failing detection jobs, flapping detection settings) came back clean.

Comparison Chart

Detection-enabled runs had a higher success rate (84.41%) than the small sample of regular runs (75.0%), and averaged far more tokens per run (41,171 vs 14,225) — consistent with detection-enabled workflows skewing toward larger, more complex agentic tasks (e.g. Ponytail Reviewer, Daily Project Performance Summary Generator at 1M+ tokens) versus the 4 regular runs, which were all lightweight smoke tests.

Misconfigured Workflows

No misconfigured workflows detected in this window.

View All Run Metrics

Regular runs (gh-aw-detection: false) — 4 total, none exceed the >3-runs-per-workflow threshold for flagging (each of the 4 belongs to a distinct workflow):

Workflow	Run	Status
Smoke Copilot	§32755907509	success
Smoke Copilot	§32766469873	success
Smoke Claude	§32766957815	success
Smoke Copilot Sub Agents	§32767410542	failure

Detection-enabled runs — 263 total across 66 workflows. Highest-volume workflows: PR Sous Chef (31 runs, 96.8% success), PR Code Quality Reviewer (25 runs, 100%), Design Decision Gate (25 runs, 84.0%), Matt Pocock Skills Reviewer (26 runs, 100%), Impeccable Skills Reviewer (23 runs, 100%), Test Quality Sentinel (22 runs, 100%), Ponytail Reviewer (24 runs, 58.3% — lowest success rate among high-volume workflows, worth a closer look independent of the detection feature itself).

All 264 detection-job invocations completed with conclusion: success — no detection-step failures were found in this window.

View Historical Trend

Two-day trend (only 2 data points so far — trending will sharpen as more days accumulate):

Date	Regular runs	Regular success %	Detection runs	Detection success %
2026-08-23	37	81.1%	321	91.3%
2026-08-24	4	75.0%	263	84.4%

Both cohorts' success rates and detection run volume declined day-over-day; regular-run volume dropped sharply (37 → 4). One day of data is too little to call this a trend rather than noise — worth revisiting after a few more snapshots.

Recommendations

• No misconfigurations to fix this cycle — detection coverage is effectively universal (98.5% of runs) and the detection job itself is healthy everywhere it ran.
• The 4 gh-aw-detection: false runs are all one-off smoke-test workflows (Smoke Copilot, Smoke Claude, Smoke Copilot Sub Agents), which is a reasonable, low-risk use of the opt-out — no action needed unless their run volume grows past the 3-run threshold.
• The day-over-day dip in both success rates (91.3%→84.4% detection, 81.1%→75.0% regular) and in detection run volume (321→263) is only two data points; keep watching via the trending history rather than reacting to a single day's dip.
• Ponytail Reviewer had the lowest success rate (58.3%) among high-volume workflows and by far the highest average token usage (182K) — unrelated to the detection feature, but worth a separate investigation given the resource cost of its failures.

References:

• §32776189539 — highest token-usage run this window (Daily Project Performance Summary Generator, 1,008,589 tokens)
• §32766584170 — Ponytail Reviewer, 959,748 tokens
• §32767410542 — only failing regular run (Smoke Copilot Sub Agents)

Generated by 🔎 Detection Analysis Report · claude · agent · 132.4 AIC · ⌖ 8.59 AIC · ⊞ 9K · ◷

• expires on Aug 27, 2026, 3:32 PM UTC-08:00

[github-actions[bot]]

👋 Just a smoke test bot passing through — dropping a haiku for the occasion:

Copilot runs deep,
Tests pass through silent commits—
Green checks bloom at dawn.

Carry on, humans! 🤖

Warning

Firewall blocked 6 domains

The following domains were blocked by the firewall during workflow execution:

• accounts.google.com
• android.clients.google.com
• clients2.google.com
• contentautofill.googleapis.com
• www.google.com
• www.gstatic.com

To allow these domains, add them to the network.allowed list in your workflow frontmatter:

network:
  allowed:
    - defaults
    - "accounts.google.com"
    - "android.clients.google.com"
    - "clients2.google.com"
    - "contentautofill.googleapis.com"
    - "www.google.com"
    - "www.gstatic.com"

See Network Configuration for more information.

📰 BREAKING: Report filed by Smoke Copilot · copilot · auto · 31.6 AIC · ⌖ 2.35 AIC · ⊞ 9.1K · ◷
_{Comment /smoke-copilot to run again
Add label smoke to run again}

[github-actions[bot]]

This discussion has been marked as outdated by Detection Analysis Report.

A newer discussion is available at Discussion #55895.