You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Window: 17:29β22:01Z (~4.5h evening cluster). A ~19.5h gap (06-14 22:02 β 06-15 17:29) is unobserved because the logs tool timed out before paginating older runs β treat counts as a partial daily sample.
Overview
Metric
Value
Completed runs
37 (32 β / 5 β) β 86.5% success
In-progress / queued
4 (incl. this audit agent)
Tokens
33.4M Β· AIC 9,963
Turns / action-min
685 / 647
Engines
copilot 29 Β· claude 10 Β· codex 2
missing_tools / missing_data / mcp_failures
0 / 0 / 0
Firewall blocked
37 / 3,283 (1.1%) β all by-design
Success rate (86.5%) is steady vs the prior day (06-14: 87.0%) and well above the 06-13 dip (68.4%). No genuine shared incident this window β the 5 failures span distinct classes across all three engines, and 3 of 5 are already-tracked recurring issues.
π Trend Charts
Workflow Health (30 days)
Success rate has held in an 85β90% band for the past week after recovering from the 06-13 trough (68%). Today's 86.5% continues that stabilization; failure volume per window remains low (5) and dominated by known classes rather than new regressions.
Token Usage (30 days)
The 7-day moving average sits near 54M tokens/day, pulled up by the 06-12 spike (~127M) which has since fallen out of the short window. Today's 33.4M reflects the partial window; per-run consumption is unremarkable with no runaway-token outliers detected.
Failures (5)
Per-failure root cause & recurrence
Workflow
Engine
Class
Root cause
Status
Repository Tree Map Generator
copilot
harness-action-timeout
Execute Copilot CLI timed out after 5 min; did partial work (tokβ580k)
π RECUR (quiet since 06-10)
Daily Safe Output Integrator
copilot
sdk tool-perm-lockout
14 permission-denied hits looped 20 min β hit action timeout
π RECUR day 9 (same wf as 06-14)
Daily Security Observability Report
copilot
agent-artifact-missing (0-tok)
setup OK (gh-aw v0.79.8) but agent emitted no output; conclusion failed Artifact not found for name: agent
π NEW
Daily Cache Strategy Analyzer
codex
gh-aw-binary-not-found-for-mcp
exit 1 Failed to find gh-aw binary for MCP serverdespite binary being built (go build) + copied to /usr/local/bin/gh-aw
π NEW signature (prior chronic fail was model-not-found)
Documentation Unbloat
claude
cache-memory-git-integrity
gitexit code 2 in check_cache_memory_git_integrity.sh β refines the prior generic "empty-output" root cause
π RECUR day 3
Note on the digest-mismatch: error line seen in conclusion logs: this is the actions/cache input echo (fail-on-cache-miss: false), not a failure cause β disregarded to avoid a false signal.
Observability insights (from logs tool)
π΄ cache-strategy 100% fail-rate β chronically broken; today a new binary-path signature.
[high] Fix Daily Cache Strategy Analyzer (codex): align the MCP-helper binary lookup path with where the build step places gh-aw. wf is at 100% fail-rate β highest ROI.
[high] Resolve copilot-SDK tool-permission lockout on Daily Safe Output Integrator (day 9; ~20 min wasted compute/run with zero output). Grant required tool perms or fix the SDK allowlist plumbing.
[medium] Documentation Unbloat: harden check_cache_memory_git_integrity.sh against the cache-memory git exit-2 (day 3 consecutive).
[watch] Daily Security Observability Report 0-tok agent-artifact-missing β first occurrence; monitor for recurrence (possible startup-failure family).
Coverage caveat
This is a partial ~4.5h sample; the mid-window gap was not retrievable within the logs-tool timeout. Recurrence/persistence counters in repo memory were incremented for the 3 recurring classes; 2 new issues registered.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
-
π Agentic Workflow Audit β 2026-06-15
Window: 17:29β22:01Z (~4.5h evening cluster). A ~19.5h gap (06-14 22:02 β 06-15 17:29) is unobserved because the logs tool timed out before paginating older runs β treat counts as a partial daily sample.
Overview
Success rate (86.5%) is steady vs the prior day (06-14: 87.0%) and well above the 06-13 dip (68.4%). No genuine shared incident this window β the 5 failures span distinct classes across all three engines, and 3 of 5 are already-tracked recurring issues.
π Trend Charts
Workflow Health (30 days)
Success rate has held in an 85β90% band for the past week after recovering from the 06-13 trough (68%). Today's 86.5% continues that stabilization; failure volume per window remains low (5) and dominated by known classes rather than new regressions.
Token Usage (30 days)
The 7-day moving average sits near 54M tokens/day, pulled up by the 06-12 spike (~127M) which has since fallen out of the short window. Today's 33.4M reflects the partial window; per-run consumption is unremarkable with no runaway-token outliers detected.
Failures (5)
Per-failure root cause & recurrence
Execute Copilot CLI timed out after 5 min; did partial work (tokβ580k)Artifact not found for name: agentFailed to find gh-aw binary for MCP serverdespite binary being built (go build) + copied to/usr/local/bin/gh-awgitexit code 2 incheck_cache_memory_git_integrity.shβ refines the prior generic "empty-output" root causeNote on the
digest-mismatch: errorline seen in conclusion logs: this is theactions/cacheinput echo (fail-on-cache-miss: false), not a failure cause β disregarded to avoid a false signal.Observability insights (from logs tool)
github-cloud.githubusercontent.comasset pulls) β by-design.Recommended actions
gh-aw. wf is at 100% fail-rate β highest ROI.check_cache_memory_git_integrity.shagainst the cache-memory git exit-2 (day 3 consecutive).Coverage caveat
This is a partial ~4.5h sample; the mid-window gap was not retrievable within the logs-tool timeout. Recurrence/persistence counters in repo memory were incremented for the 3 recurring classes; 2 new issues registered.
References: Β§27564209286 Β· Β§27572285541 Β· Β§27571281247
Beta Was this translation helpful? Give feedback.
All reactions