agentleFS
Sign inStart free
manavdharoramanavdharora/claude-agent-skills/skills/model-chat/SKILL.md

model-chat

Skill0 starsChanged 55 days agoMITView on GitHub

Spawn 5+ Claude instances into a shared conversation room where they debate, disagree, and converge on solutions. Uses round-robin turns with parallel execution within each round. Triggers on "model chat", "multi-model debate", "agent debate", "spawn a chat room", or /model-chat. Pass a topic as the argument.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Agent count: default 5, user can override

    line 21
  2. Distinctive:

    Round count: default 5, user can override

    line 22
  3. Distinctive:

    Interactive mode: after each round, prompts user for optional input to steer the debate

    line 57

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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

Model Chat

Spawn 5 Claude Sonnet instances into a shared conversation room. They debate a problem across 5 rounds using round-robin turns (all agents respond in parallel each round, see full history). A synthesizer agent then merges the debate into a final answer.

Why this works: Same model, slight framing variations = systematically different failure modes. Surfacing disagreements between instances is more valuable than any single instance's confident answer. Consensus across independent runs filters hallucinations; divergences reveal genuine judgment calls.

Execution

1. Parse the request

Extract from the user's message:

  • Topic/problem to debate
  • Mode: auto (default) or interactive (--interactive flag)
  • Agent count: default 5, user can override
  • Round count: default 5, user can override

2. Run the orchestration script

cd /Users/nicksaraev/Business && .venv/bin/python .claude/skills/model-chat/model_chat.py "<topic>"

For interactive mode:

cd /Users/nicksaraev/Business && .venv/bin/python .claude/skills/model-chat/model_chat.py "<topic>" --interactive

Optional flags:

  • --agents N — number of agents (default 5)
  • --rounds N — number of debate rounds (default 5)

3. Deliver results

The script outputs the full conversation to stdout in real-time and saves to a timestamped subdirectory:

  • active/model-chat/<YYYYMMDD-HHMMSS>/conversation.json — full structured conversation log
  • active/model-chat/<YYYYMMDD-HHMMSS>/synthesis.md — final synthesized answer
  • active/model-chat/latest — symlink to the most recent run

Present to the user:

  • Brief summary of key agreements and disagreements
  • The synthesis (or link to it)
  • Note any particularly interesting moments of debate

How it works internally

  1. 5 agents with slight framing variations (systems thinker, pragmatist, edge-case finder, UX-focused, contrarian)
  2. Round-robin: each round, all agents see full history and respond in parallel via async API calls
  3. 5 rounds of debate (25 total messages)
  4. Synthesizer: a final agent reads the entire debate and produces a structured merged answer
  5. Interactive mode: after each round, prompts user for optional input to steer the debate

Output files

Each run saves to active/model-chat/<YYYYMMDD-HHMMSS>/:

FileDescription
conversation.jsonFull structured conversation log
synthesis.mdFinal synthesized answer

A latest symlink always points to the most recent run. Previous runs are preserved.

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.

Check your setup, free