Mmcp.market

llm-council skill

by rohitg00·rohitg00/pro-workflow·2.9k stars

Provider-agnostic multi-LLM deliberation. Three phases — independent responses, cross-model anonymized ranking, chairman synthesis. Provider config from env (OPENAI/ANTHROPIC/FIREWORKS/OPENROUTER/custom OpenAI-compatible base URL). Persists transcript to a wiki page when --wiki <slug> is passed. Use when the user wants multiple AI perspectives, consensus-building, or the "LLM Council" approach for high-stakes reviews, plan critique, or contested learning rules.

A100/100content scan

Is the llm-council skill safe?

Clean: nothing in its files matched our rules. We read 2 files in the folder on 2026-09-28.

No findings.

Install the llm-council skill

A skill is a folder. Copy it into your agent's skills folder and the agent loads it when the task matches its description. 

git clone --depth 1 https://github.com/rohitg00/pro-workflow.git /tmp/pro-workflow
mkdir -p ~/.claude/skills
cp -r /tmp/pro-workflow/skills/llm-council ~/.claude/skills/llm-council
available in every project

In the Claude apps, zip the folder and upload it from the Skills settings. The folder on GitHub

The instructions your agent would load

SKILL.md as published, without the frontmatter. Read it on GitHub

LLM Council

Karpathy's LLM Council pattern, provider-agnostic. dair-academy's version hardcoded Fireworks; ours reads any OpenAI-compatible endpoint via env.

When to use

  • High-stakes plan review (/plan crosses N-file threshold)
  • Conflicting learning-rules → re-resolve via vote
  • User invokes /council "" or /wiki council
  • Architecture decisions where you want multiple viewpoints captured
  • Persisting deliberation as a wiki page for future reference

Three phases

  1. Independent: each model answers in parallel
  2. Ranking: each model ranks anonymized peer responses
  3. Synthesis: chairman model reads all responses + rankings → final answer

Provider config

Provider chosen via env. First-match wins:

Override per-run with --provider openai|anthropic|openrouter|fireworks|custom.

Default model rosters per provider live in scripts/council.js and can be overridden via --models CSV and --chairman .

Commands

node $SKILL_ROOT/scripts/council.js run "<query>" [--models id1,id2,id3] [--chairman id] [--provider <name>] [--wiki <slug>]
node $SKILL_ROOT/scripts/council.js providers
node $SKILL_ROOT/scripts/council.js show <session-id>

--wiki writes the full transcript to /derived/council/.md and registers it via wiki-cli.js page so it shows in FTS5 search.

Output

Each session writes:

~/.pro-workflow/council/<session-id>/
├── config.json           # query, models, chairman, provider
├── phase1_responses.json # raw API responses per model
├── phase2_rankings.json  # anonymized ranking outputs
├── phase3_synthesis.txt  # chairman's final answer
└── final_output.md       # human-readable bundle

Console prints the markdown bundle. Pipe to pbcopy / tee as needed.

Hard rules

  1. Never skip the ranking phase. It's the core of the council pattern.
  2. Save raw responses to disk verbatim. No summarization in storage.
  3. Anonymize responses for ranking — models see Response A/B/C/..., not peer names.
  4. The chairman sees both real names AND rankings.
  5. Display all three phases to the user. No phase elision.

Cost awareness

The script logs per-call latency + tokens on supported providers. Multiply by your provider rate to estimate. Council cost grows linearly with len(models)^2 (each model ranks all others) plus the chairman.

Default council size: 3-5 models. More models = exponentially more ranking calls.

Use with wiki

/wiki council agent-memory "should we adopt episodic memory in our agents?"

Loads agent-memory wiki context as system prompt prefix, runs council, persists transcript as wiki/derived/council/.md. The transcript becomes searchable via /wiki ask.

More skills from rohitg00/pro-workflow

  • Aagent-teamsCoordinate multiple Claude Code sessions as a team — lead + teammates with shared task lists, mailbox messaging, and file-lock claiming. Patterns for team sizing, task decomposition, and when to use teams vs sub-agents vs worktrees.
  • Aauto-setupAuto-configure quality gates, hooks, and settings for a new project. Detects project type and sets up appropriate tooling. Use when onboarding a new codebase.
  • Abatch-orchestrationDecompose large-scale changes into independent units and spawn parallel agents in isolated worktrees. Use for migrations, refactors, codemods, and any change touching 10+ files with the same pattern.
  • Cbug-captureCapture a user-reported defect as a durable GitHub issue written in the project's own domain language. Explores the codebase in parallel for context but never leaks file paths or line numbers into the issue. Use when the user reports a bug conversationally, runs a QA pass, or says "file an issue", "log this as a bug", "capture this".
  • Acompact-guardSmart context compaction with state preservation. Saves critical files, task progress, and working state before compaction, restores after. Use before manual compact or when auto-compact triggers.
  • Acontext-engineeringMaster the four operations of context engineering — Write, Select, Compress, Isolate. Manage token budgets, compaction strategies, and context partitioning to keep AI sessions sharp and efficient.
  • Acontext-optimizerOptimize token usage and context management. Use when sessions feel slow, context is degraded, or you're running out of budget.
  • Acost-trackerTrack session costs, set budget alerts, and optimize token spend. Use to check costs mid-session or set spending limits.
  • Adesign-engineeringApply interface craft when building or reviewing UI - motion, easing, timing, springs, component feel, and visual foundations. Use when building a component, animation, transition, hover or press state, modal, drawer, toast, or when polishing an interface so it feels right. Says "make this feel better", "add an animation", "polish the UI", "review this component".
  • AdeslopRemove AI-generated code slop, unnecessary comments, and over-engineering from the current branch diff. Cleans up boilerplate, simplifies abstractions, strips defensive code, and in skill-file mode lints SKILL.md files for quality. Use when cleaning up code, simplifying, removing boilerplate, before committing, or when reviewing a skill before promoting it.
  • Adomain-modelingBuild the project's shared language and bounded contexts before writing code, so names stay consistent and the agent stops paraphrasing domain concepts. Produces a CONTEXT.md glossary and decision records. Use at the start of a project or feature, or when the codebase and the people describing it speak different languages.
  • Afile-watcherConfigure file watching hooks to auto-react to config changes, env file updates, and dependency modifications. Use to set up reactive workflows.

All agent skills → · MCP servers