[copilot-cli-research] Copilot CLI Deep Research - 2026-06-03

[github-actions[bot]]

Analysis Date: 2026-06-03 · Run: §26865333861
Scope: 238 total .md workflow files, ~126 using Copilot engine

---

📊 Executive Summary

This is the first run of this research agent — a baseline for future trend tracking.

Key Findings: 5 agent files never used; 0 workflows pin engine.version; only 5 of ~126 Copilot workflows use max-continuations; engine.args/engine.harness entirely unused; 35 workflows over-permission with toolsets: [default].

Primary Recommendation: Enable max-continuations on complex long-running workflows to prevent mid-task stalls.

---

🔴 Critical Findings

1. 5 agent files defined but never used
create-safe-output-type, custom-engine-implementation, grumpy-reviewer, interactive-agent-designer, w3c-specification-writer exist in .github/agents/ but zero workflows reference them. Wire them or remove them.

2. max-continuations missing from complex workflows
Only 5 of ~126 Copilot workflows use autopilot mode. Workflows like release.md (899 lines), issue-monster.md (682 lines), agent-performance-analyzer.md (754 lines) run complex multi-step tasks without it.

engine:
  id: copilot
  max-continuations: 3  # Enable autopilot for complex tasks

🟡 Medium Priority

3. Zero version pinning — No workflow uses engine.version. Critical workflows like release.md and security-compliance.md could break on a new Copilot CLI release.

4. 35 workflows over-use toolsets: [default] — Most only need 1-2 specific toolsets. Scope down to least privilege.

5. model: small underused — 27 workflows use it, but many more lightweight reporting/summary workflows could benefit (lower cost + latency).

---

Feature Usage Matrix

Feature	Available	Used	Usage Rate
copilot-sdk	✅	62 wf	30%
engine.agent	✅	13 wf	6%
engine.bare	✅	13 wf	6%
model: small	✅	27 wf	13%
max-continuations	✅	5 wf	2%
engine.version	✅	0 wf	0%
engine.args	✅	0 wf	0%
engine.env	✅	1 wf	<1%
engine.harness	✅	0 wf	0%
engine.api-target	✅	0 wf	0%

---

7️⃣ Action Items

Immediate (this week):

• Add max-continuations: 3-5 to release.md, issue-monster.md, agent-performance-analyzer.md
• Review 5 unused agent files — wire to workflows or remove

Short-term (this month):

• Audit 35 toolsets: [default] workflows and scope down
• Pin engine.version on release.md and security-compliance.md
• Apply model: small to 5-10 additional lightweight workflows

Long-term (this quarter):

• Policy: new agent files must be wired on creation
• Add max-effective-tokens per-run limits to top-consumption workflows
• Consider BYOK for non-GitHub model experiments

---

Research Methodology: Examined pkg/workflow/copilot_engine*.go for all available features, then grep-scanned 238 .md workflow files to measure actual usage. Baseline saved to repo-memory branch for future trend tracking.

Generated by Copilot CLI Deep Research · Run §26865333861

Generated by 🔬 Copilot CLI Deep Research Agent · sonnet46 270.4K · ◷

• expires on Jun 4, 2026, 5:28 AM UTC

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-06-04T05:28:19.515Z.

Closed by Workflow