Agent Performance Report — Week of June 14, 2026

[github-actions[bot]]

Executive Summary

• Agents analyzed: 25 active workflow types (~246 total workflows, 97 runs completed today)
• Total outputs reviewed (7d): ~200 issues, 40 PRs, 127 closed (63.5% close rate)
• Average quality score: 57/100 (→ stable)
• Average effectiveness score: 55/100 (→ stable)
• AIC today: 19,985 (↓27% from Jun 12 spike; still +48% above 4-day baseline ~13,496)
• Top performers: Agentic Maintenance, Bot Detection, Auto-Triage Issues, Issue Monster, PR Sous Chef
• Critical: Code Simplifier Day 7 (api-proxy cap + HTTP 429), Dictation Prompt Generator (NEW session.idle), Smoke Gemini (NEW P1 preview model)

Analysis period: 2026-06-08 → 2026-06-14 | Run: §27499941088

---

Performance Rankings

Top Performing Agents 🏆

Rank	Agent	Quality	Effectiveness	Status
1	Agentic Maintenance	82	88	✅ 100% success
2	Bot Detection	80	90	✅ 100% success, ~13s
3	Auto-Triage Issues	80	82	✅ 100% success
4	Issue Monster	78	85	✅ 100% success, proactive
5	PR Sous Chef	78	80	✅ 13 runs today
6	PR Triage Agent	78	80	✅ 100% success
7	Claude Code User Docs Review	75	78	✅ 100% success
8	Avenger	72	75	✅ RECOVERED — 2 success runs today
9	Daily A/B Testing Advisor	75	72	✅ Active experiments
10	Package Specification Enforcer	72	75	✅ New PR today

copilot-swe-agent PR merge rate: 77% (23/30 merged in 7d) — strongest contributor.

Agents Needing Improvement 📉

Agent	Quality	Effectiveness	Primary Issue	Tracker
Code Simplifier	10	5	Day 7: HTTP 429 + api-proxy cap (maxRuns 50/50)	#39199, #39077
Dictation Prompt Generator	30	20	NEW: session.idle 570s timeout	#39196, #39200
Smoke Gemini	25	15	NEW P1: preview TTS model → unknown model	#39172
GitHub Remote MCP Auth Test	30	20	NEW: missing required tool	#39193
Daily Model Inventory Checker	35	25	Day 5+: session.idle timeout 60s	#39170
jsweep	42	35	Tool denial + 1,316 AIC/run	#39185, #39208

Inactive / Infrastructure Issues

• Smoke Trigger & Multi Caller: 100% startup_failure ([aw-failures] P2: Smoke Trigger & Smoke Multi Caller — 100% startup_failure (reusable-workflow callers), reliability blind spot #38999)
• Daily Safe Outputs Git Simulator: Day 7 streak, branch missing ([aw-failures] P1: Daily Safe Outputs Git Simulator — 5-day consecutive failure streak (memory branch missing) #39024)

Quality & Resource Analysis

Issue Close Rate by Agent Category (7d)

Category	Issues	Closed	Rate
[aw] failure notices	69	59	86% ✅
General / other	66	48	73% ✅
[aw-failures] investigations	15	5	33% ⚠️
[lint-monster] refactoring	6	1	17% ⚠️
[performance]	5	0	0% ❌
[duplicate-code]	3	0	0% ❌

Key concern: Performance and duplicate-code backlogs accumulating with 0% close rate.

PR Merge Rates (7d)

Author	PRs	Merged	Rate
copilot-swe-agent	30	23	77% ✅
app/github-actions	10	4	40% ⚠️

Avg time-to-close issues: 9.6h ✅ (fast: 11, medium: 19, slow: 0)

AIC Resource Usage

Workflow	AIC Today	Status
Contribution Check	2,002	⚠️ High (3 runs)
Daily Firewall Logs Collector	1,540	⚠️ Single-run cost
jsweep	1,316	⚠️ Tool denial + high AIC
Code Simplifier	650	❌ Fails + wastes 650 AIC daily

AIC Trend (7d): Jun 8: 13,224 → Jun 11: 10,908 (low) → Jun 12: 27,465 (spike +152%) → Jun 14: 19,985 (↓27%, still +48% above baseline)

Behavioral Patterns

Productive Patterns ✅

• Issue Monster → copilot-swe-agent pipeline: Identifies infra gaps, queues PRs (Git Simulator branch fix) — effective escalation
• Failure Investigator triage: Correctly filed 3 new P1/P2 issues today with detailed root cause analysis
• Avenger recovery: Bounced back after yesterday's failure — 2 successful runs (good resilience signal)
• copilot-swe-agent PR quality: 77% merge rate demonstrates well-scoped, high-quality contributions

Problematic Patterns ⚠️

• Session.idle timeout cluster expanding: Dictation Prompt Generator (570s) joins Daily Model Inventory Checker (60s) — may be platform-level Copilot SDK regression
• Code Simplifier AIC waste (Day 7): 650 AIC/run guaranteed-to-fail — each day accumulates waste while root fix ([perf-improvement] AIC Budget Crisis Day 5 — 6-agent cluster expanding, root fix urgently needed #39077) stalls
• app/github-actions PR gap: 40% merge rate vs 77% — agent-authored PRs need better scoping or pre-merge validation
• Refactoring backlog accumulation: lint-monster, duplicate-code, performance issues pile up with 0% action rate

Recommendations

High Priority 🔴

1. Implement Code Simplifier AIC guardrails (7 days overdue)

   • Set max-ai-credits: 2000 and max-turns: 30 — stops ~650 AIC/day waste
   • Tracked: [perf-improvement] AIC Budget Crisis Day 5 — 6-agent cluster expanding, root fix urgently needed #39077 (root fix), [aw-failures] P1: Code Simplifier exhausts api-proxy invocation cap (maxRuns 50/50) → CAPIError 429, 6/6 fail + unclassified #39199 (api-proxy escalation)

2. Investigate Copilot SDK session.idle timeout regression

   • Two agents now failing with identical pattern (Dictation: 570s, Model Inventory: 60s)
   • Check SDK version/deployment; no systemic tracker exists yet

3. Fix Smoke Gemini model reference ([aw-failures] P1: Smoke Gemini uses TTS preview model gemini-3.1-flash-tts-preview → unknown_model_ai_credits, 100% agent fail [Content truncated due to length] #39172) — swap gemini-3.1-flash-tts-preview for stable ID

Medium Priority 🟡

4. Improve app/github-actions PR merge rate (40% → target 70%+) — audit 6 open PRs for common rejection causes
5. Manage refactoring backlog — pause new lint/duplicate/performance issues until existing ones are actioned, or add auto-assignment
6. Optimize jsweep AIC ([agentic-token-optimizer] Optimization: jsweep - JavaScript Unbloater (58 turns, 1315 AIC/run) #39208) — apply optimizer recommendations to cap ≤500 AIC/run

Low Priority 🟢

7. Fix GitHub Remote MCP Auth Test config drift ([aw] GitHub Remote MCP Authentication Test is missing required tool #39193)
8. Review Smoke Trigger startup_failure pattern ([aw-failures] P2: Smoke Trigger & Smoke Multi Caller — 100% startup_failure (reusable-workflow callers), reliability blind spot #38999, 2-week blocker)

Trends

Metric	This Week	Last Week	Δ
Quality score	57/100	62/100	↓5
Effectiveness score	55/100	58/100	↓3
copilot-swe-agent merge rate	77%	~70%	↑7% ✅
Avg time-to-close	9.6h	~11h	↓1.4h ✅
AIC daily	~19,985	~13,500	↑48% ⚠️
Active P1 blockers	5	4	↑1

Actions Taken This Run

• Updated agent-performance-latest.md and shared-alerts.md in shared memory
• No new issues filed — all actionable items already tracked today by Failure Investigator ([aw-failures] P1: Code Simplifier exhausts api-proxy invocation cap (maxRuns 50/50) → CAPIError 429, 6/6 fail + unclassified #39199, [aw-failures] P2: Dictation Prompt Generator false-failure — Copilot session.idle 570s timeout after PR #39195 already created #39200, [aw-failures] P1: Smoke Gemini uses TTS preview model gemini-3.1-flash-tts-preview → unknown_model_ai_credits, 100% agent fail [Content truncated due to length] #39172)

Next Steps

1. 🔴 Urgent: Ship Code Simplifier AIC guardrails ([perf-improvement] AIC Budget Crisis Day 5 — 6-agent cluster expanding, root fix urgently needed #39077) — 7 days of compounding waste
2. 🔴 Urgent: Investigate Copilot SDK session.idle timeout as platform regression
3. 🟡 48h: Fix Smoke Gemini model reference ([aw-failures] P1: Smoke Gemini uses TTS preview model gemini-3.1-flash-tts-preview → unknown_model_ai_credits, 100% agent fail [Content truncated due to length] #39172)
4. 🟡 1 week: Audit app/github-actions open PRs for merge rate improvement
5. 🟡 1 week: Decide backlog management strategy for refactoring/performance issues

---

Analysis period: 2026-06-08 to 2026-06-14 | Next report: 2026-06-15
Run: §27499941088

References:

• §27499941088 — This run
• §27467672212 — Yesterday's Agent Performance run
• AIC Audit: [agentic-token-audit] Daily AIC Usage Audit — 2026-06-14 #39204

Generated by ⚡ Agent Performance Analyzer - Meta-Orchestrator · 362.7 AIC · ⌖ 20.3 AIC · ⊞ 21.8K · ◷

• expires on Jun 15, 2026, 5:27 AM UTC-08:00

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-06-15T13:27:29.623Z.

Closed by Workflow