[copilot-cli-research] Copilot CLI Deep Research - 2026-06-28

[github-actions[bot]]

Analysis Date: 2026-06-28 · Scope: 257 total workflows, 126 using Copilot (49%)
Run: §28312118143

---

📊 Executive Summary

10 missed opportunities identified across 126 Copilot workflows, with 3 critical gaps showing 0% adoption after 12+ consecutive research runs. The Copilot SDK mode has achieved 51% adoption (64/126) — yet every single SDK workflow is missing the max-tool-denials safety net that prevents infinite tool-denial loops.

---

🔴 Critical Findings (High Priority)

1. max-tool-denials — 0/64 SDK Workflows (12+ Research Runs at Zero)

The max-tool-denials field stops Copilot SDK inference when tools are repeatedly denied, preventing runaway loops that burn AI credits. Zero of 64 copilot-sdk: true workflows configures this. The default is 5; recommended is 8 for complex workflows.

engine:
  id: copilot
  copilot-sdk: true
max-tool-denials: 8

Affected: craft.md, daily-safe-output-integrator.md, architecture-guardian.md, agent-performance-analyzer.md (all 64 SDK workflows).

2. tools.timeout / tools.startup-timeout — Near Zero (1/126)

Only 1 of 126 workflows uses per-tool timeout or MCP startup timeout. Slow/unresponsive MCP servers stall silently until timeout-minutes fires. This makes failures harder to debug and wastes runner time.

tools:
  timeout: 120          # 120s per tool call
  startup-timeout: 60   # 60s for MCP server startup
  github:
    mode: gh-proxy

3. network.blocked — 0 Workflows (Never Used)

Zero workflows use the network.blocked field for domain exclusions. Security-sensitive write workflows like release.md, stale-pr-cleanup.md, and code-scanning-fixer.md would benefit from explicit domain blocklists.

---

🟡 Medium Priority Opportunities

#	Opportunity	Current	Impact
4	max-continuations for long tasks	8/126 (6%)	Multi-pass completion for security/spec workflows
5	Wire 5 unused agent files	5 unused	grumpy-reviewer, w3c-specification-writer, create-safe-output-type, etc.
6	engine.version pinning in release workflows	9/126 (7%)	Reproducibility in release.md, dependabot-burner.md
7	web-fetch regression	26→12 workflows	Likely inadvertent removal in recent refactors

🟢 Low Priority Opportunities

#	Opportunity	Notes
8	engine.api-target for GHEC/GHES	0 usage; document for enterprise users
9	Custom copilot-sdk-driver	0 usage; add reference implementation
10	LSP expansion to Go workflows	3/126; refiner.md, tidy.md would benefit

---

📈 Usage Statistics

Feature	Count	%	Trend
copilot-sdk: true	64/126	51%	✅ +156% since May-26
tools.cli-proxy	112/126	89%	✅ Strong
tools.github (MCP)	86/126	68%	✅ Strong
tools.cache-memory	44/126	35%	✅ Growing
copilot-requests: write	53/126	42%	✅ Growing
tools.repo-memory	21/126	17%	✅ Growing
max-continuations	8/126	6%	⚠️ Underused
max-tool-denials	0/64 SDK	0%	🔴 Critical gap
tools.timeout	~1/126	1%	🔴 Critical gap
network.blocked	0/126	0%	🔴 Never used

---

7️⃣ Action Items

Immediate (this week):

• Add max-tool-denials: 8 to all 64 copilot-sdk: true workflows
• Add tools.startup-timeout: 60 to workflows with MCP servers
• Investigate web-fetch regression (26→12 workflows)

Short-term (this month):

• Enable max-continuations: 2-3 in security-compliance.md, weekly-blog-post-writer.md, spec-extractor.md
• Pin engine.version in release.md and dependabot-burner.md
• Wire or remove 5 unused .github/agents/ files

Long-term (this quarter):

• Expand lsp: to Go editing workflows (refiner.md, tidy.md)
• Create reference copilot-sdk-driver implementation
• Add network.blocked to 3+ write-sensitive workflows

---

Trend History (May 2026 → June 2026)

Metric	May-21	Jun-26	Jun-28	Status
Total workflows	233	252	257	+24
Copilot workflows	100	128	126	+26
copilot-sdk	~25	63	64	+156%
max-tool-denials	0	0	0	🔴
network.blocked	0	0	0	🔴
tools.timeout	0	~1	~1	🔴
model overrides	15	70	~70	✅
engine.harness	0	1	1	🆕 first use

Generated by 🔬 Copilot CLI Deep Research Agent · 141.8 AIC · ⌖ 25.2 AIC · ⊞ 7.4K · ◷

• expires on Jun 28, 2026, 9:19 PM UTC-08:00

[pelikhan]

/q this AW should create an issue instead of discussion

[github-actions[bot]]

🔧 Pay attention, 007! Q is preparing your gadgets for this discussion comment...

[pelikhan]

/plan

[github-actions[bot]]

🚀 Agentic Commands has started processing this discussion comment

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-06-29T05:19:13.009Z.

Closed by Workflow