[External Plugin]: agents-system-setup

[ytthuan]

Plugin name

agents-system-setup

Short description

Bootstrap, update, improve, or replicate compact multi-agent systems across GitHub Copilot CLI, Claude Code, OpenCode, OpenAI Codex, and Gemini CLI. Generates AGENTS.md governance, specialized subagents, and approval-gated plugin/MCP recommendations.

GitHub repository

ytthuan/agents-system-setup

Plugin path inside the repository

plugins/agents-system-setup

Ref to review

v1.11.0

Commit SHA to review

b19eb8f3cdf6d2ec1a737b6cbd8c47cbafe53980

Version

1.11.0

License identifier

MIT

Author name

ytthuan

Author URL

https://github.com/ytthuan

Homepage URL

https://github.com/ytthuan/agents-system-setup

Keywords

agents, copilot-cli, claude-code, opencode, codex, gemini-cli, subagents, agent-teams, scaffolding, mcp

Additional notes for reviewers

Resubmitting at v1.11.0 after addressing the prior review note about overlap with /init and native AGENTS.md interop.

This plugin complements /init, it does not duplicate it. /init seeds one Copilot-only guidance file (a valid input to this plugin's update/improve mode). Beyond that, agents-system-setup provides uplift no built-in command offers:

• Cross-runtime replication via a Canonical IR — generates and ports the same agent system across five runtimes (Copilot CLI, Claude Code, OpenCode, OpenAI Codex, Gemini CLI), each in its native agent/skill/permission format. /init is Copilot-only, single-file.
• A governed multi-agent system, not one file — Agent Roster, Capability Matrix, Security & Audit Matrix, Threat Model, Architecture/ADR decisions, and N specialized subagents with parallel-safety waves under a host-session Orchestration Operating Model.
• Lifecycle modes beyond bootstrap — update (non-destructive managed blocks), improve (audit → opt-in fixes), upgrade (version-stamped migration), replicate (port between runtimes).
• A mandatory MCP approval gate (no MCP config written without explicit approval) + a source-backed governance baseline (OWASP GenAI, NIST SSDF, MCP Security Best Practices, SLSA).

A new README section, "Relationship to Copilot /init and native AGENTS.md support", makes this boundary explicit. v1.11.0 also adds cross-session orchestration (GitHub Copilot app /orchestrate + create_session) as an advisory parallelism primitive. Public repo, MIT, immutable tag v1.11.0 + full SHA. Self-contained plugin at plugins/agents-system-setup (root plugin.json + skills/).

Submission checklist

• The plugin lives in a public GitHub repository.
• The ref and/or sha I provided is immutable (release tag and/or full 40-character commit SHA), not a branch.
• This submission follows this repository's contribution, security, and responsible AI policies.
• This plugin is not already listed in the Awesome Copilot marketplace.

[github-actions]

✅ External plugin intake passed

This submission passed automated intake validation and quality checks and is ready for maintainer review.

• Plugin: agents-system-setup
• Repository: https://github.com/ytthuan/agents-system-setup
• Ref: v1.11.0
• SHA: b19eb8f3cdf6d2ec1a737b6cbd8c47cbafe53980

Quality gate summary

Gate	Status
skill-validator	pass
install smoke test	pass

• skill-validator: pass
• install smoke test: pass
• overall: pass

skill-validator output

Found 1 skill(s)
[agents-system-setup] 📊 agents-system-setup: 18,793 BPE tokens [chars/4: 20,160] (comprehensive ✗), 26 sections, 0 code blocks
[agents-system-setup]    ⚠  Skill is 18,793 BPE tokens (chars/4 estimate: 20,160) — "comprehensive" skills hurt performance by 2.9pp on average. Consider splitting into 2–3 focused skills.
[agents-system-setup]    ⚠  No code blocks — agents perform better with concrete snippets and commands.
✅ All checks passed (1 skill(s))

Install smoke test output

Install smoke test succeeded. Verified /tmp/external-plugin-quality-hlEq2s/copilot-home/.copilot/installed-plugins/external-plugin-intake/agents-system-setup/plugin.json.

Canonical external.json payload

{
  "name": "agents-system-setup",
  "description": "Bootstrap, update, improve, or replicate compact multi-agent systems across GitHub Copilot CLI, Claude Code, OpenCode, OpenAI Codex, and Gemini CLI. Generates AGENTS.md governance, specialized subagents, and approval-gated plugin/MCP recommendations.",
  "version": "1.11.0",
  "author": {
    "name": "ytthuan",
    "url": "https://github.com/ytthuan"
  },
  "repository": "https://github.com/ytthuan/agents-system-setup",
  "homepage": "https://github.com/ytthuan/agents-system-setup",
  "license": "MIT",
  "keywords": [
    "agents",
    "copilot-cli",
    "claude-code",
    "opencode",
    "codex",
    "gemini-cli",
    "subagents",
    "agent-teams",
    "scaffolding",
    "mcp"
  ],
  "source": {
    "source": "github",
    "repo": "ytthuan/agents-system-setup",
    "path": "plugins/agents-system-setup",
    "ref": "v1.11.0",
    "sha": "b19eb8f3cdf6d2ec1a737b6cbd8c47cbafe53980"
  }
}

View workflow run

[ytthuan]

/rerun-intake

[aaronpowell]

/reject Copilot already has a built-in command for generating an initial file for the repo, /init, as well as having interop support for files like AGENTS.md and fallback support for files used by other agentic coding tools (where applicable).

[github-actions]

❌ External plugin rejected

A maintainer rejected agents-system-setup, and the submission issue has been closed.

Reason

Copilot already has a built-in command for generating an initial file for the repo, /init, as well as having interop support for files like AGENTS.md and fallback support for files used by other agentic coding tools (where applicable).

If you address the feedback, edit this issue with the updated details and have the issue author or a maintainer comment /rerun-intake to re-run automated intake.

[ytthuan]

Thank you for taking the time to review this, and for the clear reason — that's fair, the headline artifact really is an AGENTS.md, which /init already seeds and Copilot reads natively. I've updated the submission to v1.11.0 to address the feedback, and I'd like to respectfully clarify the scope, since I think the AGENTS.md file is the table-stakes part rather than the value.

agents-system-setup is meant to complement /init, not replace it — an /init-seeded AGENTS.md is actually a valid input to the plugin's update/improve mode. What it adds beyond /init + native interop:

1. Cross-runtime replication via a Canonical IR — it generates and ports the same agent system across five runtimes (Copilot CLI, Claude Code, OpenCode, OpenAI Codex, Gemini CLI), each in its native agent/skill/permission format. /init is Copilot-only and single-file.
2. A governed multi-agent system, not one guidance file — Agent Roster, Capability Matrix, Security & Audit Matrix, Threat Model, Architecture/ADR decisions, and N specialized subagents with parallel-safety waves under a host-session Orchestration Operating Model.
3. Lifecycle modes beyond bootstrap — update (non-destructive managed blocks), improve (audit an existing system → opt-in fixes), upgrade (version-stamped migration), and replicate (port between runtimes).
4. Security gating — a mandatory MCP approval gate (no MCP config is written without explicit approval) plus a source-backed baseline (OWASP GenAI, NIST SSDF, MCP Security Best Practices, SLSA).

To make this boundary explicit rather than implicit, I added a "Relationship to Copilot /init and native AGENTS.md support" section to the README: https://github.com/ytthuan/agents-system-setup#relationship-to-copilot-init-and-native-agentsmd-support

If the multi-runtime + governance/lifecycle scope reads as meaningful uplift, I'd be grateful for another look. If it's still outside what the collection is looking for, I completely understand and appreciate the consideration. Thanks again.

[ytthuan]

/rerun-intake