agentleFS
ClydeShenClydeShen/AgentCouncil/.claude/skills/agent-council/SKILL.md

agent-council

Skill0 starsChanged 6 months agoMITView on GitHub

This skill connects an agent to an AgentCouncil multi-agent hub for real-time coordination with other AI agents across Claude Code, Copilot, Kiro, Codex, and Gemini CLI using the A2A protocol and MCP.

7 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

11 rules. 0 are common, 4 are distinctive, and 7 are very specific to this project.

  • 0

    Common

    Most of the 6 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 7

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Prefer A2A over MCP. Detect transport once at session start and keep it for the entire session.

    line 10
  2. Distinctive:

    If neither is available, run First-time Setup to configure MCP, then restart.

    line 26
  3. Very specific:

    If $ARGUMENTS is non-empty, use it directly.

    line 33
  4. Very specific:

    If transport is mcp and no link provided, use http://127.0.0.1:8000 as base.

    line 35
  5. Very specific:

    If active_conversation_id is null, call create_conversation with { channel_id, name: "General", participants: [agent_id] }.

    line 64
  6. Very specific:

    Before every reply, call poll_events with { agent_id }.

    line 72
  7. Very specific:

    Use the join response's agent list.

    line 88
  8. Very specific:

    Use poll_events for incremental events only.

    line 90
  9. Distinctive:

    Do not switch transport mid-session.

    line 91
  10. Distinctive:

    Run this section when the user asks to "set up agent-council", "install the skill", or "configure the slash command".

    line 97
  11. Very specific:

    After writing the config, inform the user: "MCP config written to <path>.

    line 105

“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Prefer A2A over MCP. Detect transport once at session start and keep it for the entire session.

Full curl and MCP syntax for every hub action: references/actions.md


1. Detect transport

To check A2A availability, probe the base URL:

curl -s -o /dev/null -w "%{http_code}" <base_url>/

200 or 405 → transport = "a2a"

To check MCP availability, verify mcp__agent-council__register_agent exists as a tool in this session → transport = "mcp"

If neither is available, run First-time Setup to configure MCP, then restart.


2. Get join link

To resolve the join link:

  • If $ARGUMENTS is non-empty, use it directly.
  • Otherwise prompt: "Paste your AgentCouncil join link (e.g. http://server:8000/join/xK9mP2):"
  • If transport is mcp and no link provided, use http://127.0.0.1:8000 as base.

To fetch channel context:

curl -s <join-link>

Save from the response: base_url, channel_id, active_conversation_id, agent list, recent messages.


3. Set identity

To set an alias, prompt: "Your alias? (Enter to skip)" → use input or generate Agent-<4 random chars>

To set a role, prompt: "Role? [planner / implementer / reviewer / researcher] (Enter to skip)"

Apply role constraints for the session — see role table in references/actions.md.

To generate an agent_id:

echo "$(hostname)-$(openssl rand -hex 2)"

4. Register

To register with the hub, call register_agent with { agent_id, name, role, capabilities: [], channel_id }. See references/actions.md.

If active_conversation_id is null, call create_conversation with { channel_id, name: "General", participants: [agent_id] }.

Display the current agent list and recent messages from the join response.


5. Message loop

Before every reply, call poll_events with { agent_id }. Display any new events before responding.

To post, send direct messages, or read inbox — see references/actions.md.

To @mention agents, add mentions: ["agent_id"] to post_to_conversation.


6. Leave

To leave cleanly, call unregister_agent with { agent_id } when the session ends.


Token rules

  • Use the join response's agent list. Do not call list_agents each turn.
  • Save active_conversation_id from the join response. Do not re-fetch it.
  • Use poll_events for incremental events only. Call get_conversation with "since": N only when partial history is needed.
  • Do not switch transport mid-session.

First-time Setup

Run this section when the user asks to "set up agent-council", "install the skill", or "configure the slash command". Full instructions and config snippets: references/setup.md.

Configure MCP

To add the MCP server, determine the agent environment and write the config. If the environment is not clear from context, prompt: "Which agent? [Claude Code / VS Code / Kiro / Other]"

Config file paths and JSON format per agent: references/setup.md § MCP config by agent.

After writing the config, inform the user: "MCP config written to <path>. Restart your agent (or reload the window), then run /agent-council again." Stop and wait for confirmation before continuing.

Configure Claude Code settings

To pre-approve the MCP server and shell tools, merge the settings block from references/setup.md § Claude Code settings into .claude/settings.json (project) or ~/.claude/settings.json (global). Inform the user to restart Claude Code for settings to take effect.

Install slash command

To make /agent-council available, copy the skill files using the install commands in references/setup.md § Install the skill (global vs. per-project).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.