Problem
The Metrics Collector's shared snapshot at /tmp/gh-aw/repo-memory/default/metrics/latest.json (timestamp 2026-08-05T03:38:39Z) reports safe_outputs: {issues_created: 0, prs_created: 0, comments_added: 0, discussions_created: 0} for all 31 active workflows, and ecosystem.total_safe_outputs: 0. The collection_note field explains why:
"agenticworkflows logs tool timed out for full -1d window (>60s at count>=100); data limited to most recent 50 runs (partial 24h window, back to run_id 30965246877). Safe-output/engagement counts unavailable from this data source in this run."
The same truncated-window symptom appears in the prior day's snapshot (metrics/daily/2026-08-04.json): "agenticworkflows logs tool timed out for full -1d window (>60s at count>=100); data limited to most recent 50 runs (~3.5h window...). Safe-output counts ... were not available."
Impact
Ground-truth gh issue list/gh api search/issues for the same window shows dozens of real agent-created issues/comments (e.g. #50506, #50516, #50515, #50502, #50519, #50521), so the zeroed safe_outputs figures are false negatives from a data-collection timeout, not evidence of an ecosystem-wide output freeze. This blocks any downstream quality/effectiveness/engagement scoring that depends on safe_outputs or engagement fields (used by Agent Performance Analyzer, Campaign Manager, Workflow Health Manager) — the last two collection runs in a row have hit this same truncation.
Suggested Fix
The agenticworkflows logs tool call inside the Metrics Collector workflow appears to time out (>60s) whenever it tries to page beyond the first ~50 runs or a count>=100 request in the -1d window. Either:
- Increase the timeout/retry budget for that specific tool call, or
- Paginate in smaller batches (e.g. 20-run pages) with partial-result accumulation instead of a single large request, or
- Fall back to a lighter read path (e.g.
gh api search across created issues/PRs/comments) specifically for the safe_outputs/engagement fields when the primary logs-based path times out, rather than zeroing those fields silently.
Suggested Agent
Metrics Collector workflow owner / Copilot SWE Agent.
Estimated Effort
Low–Medium (pagination or timeout/retry tuning in the collector's data-fetch step).
Generated by ⚡ Agent Performance Analyzer - Meta-Orchestrator · auto · 117.6 AIC · ⌖ 7.84 AIC · ⊞ 17K · ◷
Problem
The Metrics Collector's shared snapshot at
/tmp/gh-aw/repo-memory/default/metrics/latest.json(timestamp 2026-08-05T03:38:39Z) reportssafe_outputs: {issues_created: 0, prs_created: 0, comments_added: 0, discussions_created: 0}for all 31 active workflows, andecosystem.total_safe_outputs: 0. Thecollection_notefield explains why:The same truncated-window symptom appears in the prior day's snapshot (
metrics/daily/2026-08-04.json): "agenticworkflows logs tool timed out for full -1d window (>60s at count>=100); data limited to most recent 50 runs (~3.5h window...). Safe-output counts ... were not available."Impact
Ground-truth
gh issue list/gh api search/issuesfor the same window shows dozens of real agent-created issues/comments (e.g. #50506, #50516, #50515, #50502, #50519, #50521), so the zeroedsafe_outputsfigures are false negatives from a data-collection timeout, not evidence of an ecosystem-wide output freeze. This blocks any downstream quality/effectiveness/engagement scoring that depends onsafe_outputsorengagementfields (used by Agent Performance Analyzer, Campaign Manager, Workflow Health Manager) — the last two collection runs in a row have hit this same truncation.Suggested Fix
The
agenticworkflowslogs tool call inside the Metrics Collector workflow appears to time out (>60s) whenever it tries to page beyond the first ~50 runs or acount>=100request in the -1d window. Either:gh apisearch across created issues/PRs/comments) specifically for thesafe_outputs/engagementfields when the primary logs-based path times out, rather than zeroing those fields silently.Suggested Agent
Metrics Collector workflow owner / Copilot SWE Agent.
Estimated Effort
Low–Medium (pagination or timeout/retry tuning in the collector's data-fetch step).