[copilot-cli-research] Copilot CLI Deep Research - 2026-06-24

[github-actions[bot]]

Analysis Date: 2026-06-24 | Run: §28076257791
Scope: 251 total workflows, 121 Copilot (48%), 63 Copilot SDK (52% of Copilot)

---

📊 Executive Summary

Key Findings: 8 persistent zero-usage gaps; min-integrity regression −8 (43→35); max-tool-denials absent from all 63 SDK workflows; MCP session/tool timeouts never deployed despite 7-week availability.

Primary Recommendation: Add max-tool-denials: 8 to all 63 SDK workflows — a one-line change that prevents runaway tool-denial credit burns.

---

Critical Findings

🔴 High Priority

Issue	Affected
max-tool-denials absent from ALL 63 SDK workflows	63/63 (0%)
min-integrity regression (43→35, −8 since Jun-22)	−8 workflows
mcp.session-timeout / mcp.tool-timeout never used (7 weeks old!)	12 workflows ≥60 min

🟡 Medium Priority

Opportunity	Current Usage
max-continuations autopilot	7/121 copilot (6%)
engine.agent custom agents	8/121 (7%); 5 unused agent files
engine.version pinning	0 genuine engine pins
startup-timeout for MCP tools	1/251 (0.4%)

---

2️⃣ Feature Usage Matrix

Feature	Used	Trend vs Jun-22
engine.agent (custom)	8/121	Stable
engine.args	0	❌ 10+ runs
engine.api-target	0	❌ All runs
engine.harness	0	❌ All runs
engine.version (engine pin)	0	❌ All runs
engine.driver (SDK)	6/63	+3 ✅
mcp.session-timeout	0	❌ Since May launch
mcp.tool-timeout	0	❌ Since May launch
max-tool-denials	0/63 SDK	❌ 10+ runs
min-integrity	35 (14%)	⚠️ −8 regression
blocked-domains	0	❌ All runs
max-continuations	7/121	Stable
max-daily-ai-credits	73/251	✅ Stable
cache-memory	72/251	✅ Stable
tracker-id	90/251	+1 ✅

---

3️⃣ Top Missed Opportunities

🔴 HIGH: max-tool-denials for SDK Workflows

copilot-sdk: true
max-tool-denials: 8   # Add to all 63 SDK workflows

Prevents infinite tool-denial loops from burning credits.

🔴 HIGH: Restore min-integrity Coverage

activation:
  min-integrity: approved  # For PR-triggered workflows

8 workflows lost this protection since June 22.

🔴 HIGH: mcp.session-timeout for Long-Running Workflows

engine:
  id: copilot
  mcp:
    session-timeout: 4h    # 12 workflows run 60+ minutes
    tool-timeout: 5m

agent-persona-explorer.md (180 min), linter-miner.md (120 min) are prime candidates.

🟡 MEDIUM: Expand max-continuations

engine:
  id: copilot
  max-continuations: 10  # 40 long-timeout workflows have none

🟡 MEDIUM: Pin engine.version for Smoke Tests

engine:
  id: copilot
  version: "0.0.422"  # Zero genuine pins today

---

4️⃣ Historical Trends (5 runs)

Metric	May-21	Jun-22	Jun-24	Trend
Total workflows	233	249	251	📈
Copilot	100	123	121	↔ Plateau
cache-memory	30	72	72	✅ Plateau
max-daily-ai-credits	0	73	73	✅ Stable
min-integrity	—	43	35	⚠️ −8
max-tool-denials	0	0	0	❌ Persistent
mcp.session-timeout	0	0	0	❌ Persistent
engine.args	2	0	0	⚠️ Regressed

---

5️⃣ Action Items

Immediate (this week):

• Add max-tool-denials: 8 to all 63 SDK workflows
• Investigate min-integrity regression (which 8 workflows lost it?)

Short-term (this month):

• Add engine.mcp.session-timeout: 4h to 12 workflows with 60+ min timeouts
• Add max-continuations: 10 to agent-persona-explorer.md (180 min) and linter-miner.md (120 min)
• Wire or remove 5 unused .github/agents/ files

Long-term (this quarter):

• Pin engine.version for smoke tests
• Review min-integrity strategy — only 14% coverage across 251 workflows
• Add tools.startup-timeout: 120 to reduce multi-MCP flakiness

---

Persistent Gaps (all runs, never resolved):
engine.args, engine.api-target, engine.harness, engine.token-weights, mcp.session-timeout, mcp.tool-timeout, max-tool-denials (SDK), blocked-domains

Generated by Copilot CLI Deep Research (Run: §28076257791)

Generated by 🔬 Copilot CLI Deep Research Agent · 240.8 AIC · ⌖ 13.8 AIC · ⊞ 7.4K · ◷

• expires on Jun 24, 2026, 9:14 PM UTC-08:00

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-06-25T05:14:35.640Z.

Closed by Workflow