[Outcome Report] Safe Output Outcomes Report — 2026-06-16

[github-actions]

Caution

agentic threat detected
Threat detection flagged this output in warn mode. Manual review is REQUIRED before any follow-up automation.

Details

The threat detection results could not be parsed.

Review the workflow run logs for details.

Workflow Health — 2026-06-16

Executive read: Acceptance rate is strong (100% of completed items), but two workflows (Issue Monster, PR Sous Chef) are stuck with high pending volume (19/25 total). Eleven workflows generate only "unknown" outcomes with no acceptance/rejection signal — these need dedicated evaluators or clearer acceptance criteria. Fallback evaluations (20.7% of items) indicate moderate signal quality.

Workflow	Status	Lifecycle	References
Issue Monster	🟩🟩🟩🟩🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨	🔴 stuck	🟩 #39412 · 🟩 #39451 · 🟩 #39443 · 🟩 #39412 · 🟨 #39413 · 🟨 #39441 · 🟨 #39443 · 🟨 #39459 · 🟨 #39451 · 🟨 #39414 · 🟨 #39415 · 🟨 #39452
PR Sous Chef	🟩🟩🟨🟨🟨🟨🟨	🔴 stuck	🟩 #39430 · 🟩 #39430 · 🟨 #39100 · 🟨 #39457 · 🟨 #39430 · 🟨 #39300 · 🟨 #39386
Smoke CI	🟩🟩🟩🟩🟩🟩🟩🟨	🟡 in flight	🟩 #39467 · 🟩 #39467 · 🟩 #39466 · 🟩 #39466 · 🟩 #39466 · 🟩 #39450 · 🟩 #39450 · 🟨 #39466
PR Description Updater	🟩🟩🟩	🟢 resolving	🟩 #39467 · 🟩 #39466 · 🟩 #39450
[aw] Failure Investigator (6h)	🟩🟩	🟢 resolving	🟩 #39451 · 🟩 #39452
Semantic Function Refactoring	🟩🟨	🟡 in flight	🟩 #39298 · 🟨 #39468
Daily Reliability Review	🟨	🟡 in flight	🟨 #39465
Daily Documentation Healer	🟨	🟡 in flight	🟨 #39472
Daily Caveman Optimizer	🟨	🟡 in flight	🟨 #39457
Daily Ambient Context Optimizer	🟨	🟡 in flight	🟨 #39453
Daily Safe Output Tool Optimizer	🟩	🟢 resolving	🟩 #39459
Contribution Check	🟩	🟢 resolving	🟩 #39430
Underdefined (11 workflows)	⬜	⚪	Daily Regulatory Report, Sentrux Report, MCP Inspector, others

Legend: 🟩 accepted · 🟥 rejected · 🟨 pending · ⬜ unknown | 🟢 resolving · 🟡 in flight · 🟠 aging · 🔴 stuck · ⚪ underdefined

🔴 Action Items

1. Stuck workflows — Issue Monster (14 pending) and PR Sous Chef (5 pending) need resolution. Review if items are awaiting timeouts, blocked externally, or prompt-driven over-generation.

2. Underdefined workflows (11 total) — These only produce "unknown" outcomes. Assign human evaluators or refine safe-output types to use noop for non-actionable reports.

3. High fallback rate (20.7%) — 12 items evaluated by existence only (weak signal). Add stronger evidence: engagement signals, merged PRs, resolved issues, or human approval.

4. Zero-touch rate (0%) — All accepted items show no human follow-up beyond bot actions. Tighten acceptance criteria or switch low-engagement workflows to noop.

Metrics Summary

Metric	Value	Status
Acceptance rate	100.0%	🟢 strong (21 of 21 completed)
Pending items	25	needs resolution
Runs checked	32	—
Strong evidence	7 items	merged/approved
Medium evidence	14 items	engaged/retained
Weak evidence (fallback)	12 items	existence only
Median resolution time	13 minutes	—

📊 Measured by Outcome Collector · 72.2 AIC · ⌖ 4.6 AIC · ⊞ 23.7K

• expires on Jun 22, 2026, 5:14 PM UTC-08:00