Repository navigation
[detection-analysis] Detection Analysis Report — 2026-10-31 #73579
Closed
Replies: 1 comment
|
This discussion has been marked as outdated by Detection Analysis Report. A newer discussion is available at Discussion #73906. |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Summary
aw_info.json/run_summary.json, likely truncated by the storage budget) and are excluded from classification; coverage for those is unknowngh-aw-detection: false): 54 (9.8%)Note
No detection-related misconfigurations were detected in this window. All workflows explicitly disabling detection had ≤3 runs in this 24h window (rule 1 threshold not met), no run showed the
detectionjob itself failing while detection was enabled (rule 2), and no workflow toggled the flag mid-window (rule 3).Comparison Chart
Detection-enabled runs show a lower success rate and roughly 2x the average token usage compared to regular runs. This is directionally expected since detection adds a dedicated job/step (and most production workflows use detection by default), but the gap is worth tracking over time via the trend chart below.
Misconfigured Workflows
No misconfigured workflows detected in this window.
View All Run Metrics
203 workflows observed. "Detection" = all runs had detection enabled (default or explicit
true); "Regular" = all runs hadgh-aw-detection: false; "Mixed" = alternated (none observed this window).View Historical Trend
Trending history now holds 32 distinct days (2026-08-23 → 2026-10-31, with some gaps) in
/tmp/gh-aw/cache-memory/trending/detection-metrics/history.jsonl. Today's data point was appended.Recommendations
gh-aw-detection: false), rule 2 (detection job itself failing), or rule 3 (flag toggling mid-window).agentjob succeed and 129/130 had thedetectionjob succeed, yet the overall run conclusion wasfailurein 127/130 cases. The failure is attributable to a later job (e.g.safe_outputs) in at least one sampled run, not to detection — worth a dedicated investigation outside this report's scope.aw_info.json/run_summary.json(likely truncated by the storage budget) and were excluded from detection classification. If a future run needs tighter precision on the detection/regular split, increase the log-collection storage budget or target a narrower time window to avoid gaps.All reactions