Skip to main content
affpapa.org
RU EN
skill CX Codex

llm-council · am-will — Claude Code skill: a council of CLI planners

This skill runs several CLI agents as independent planners, strips their identities, and merges the results into one final plan. Useful when a single model's bias is the problem.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓ 
npx skills add am-will/codex-skills --skill llm-council -a claude-code
Open on GitHub

What llm-council does

The skill spins up a council of planners: Codex, Claude, Gemini, OpenCode, or your own command. Each agent gets the same brief and writes a plan without seeing the others. Then the plans are anonymized, shuffled, and handed to a judge, so neither model name nor position sways the verdict.

You run it with python3 scripts/llm_council.py run --spec /path/to/spec.json. The lineup lives in agents.planners, and agents.judge overrides the judge; leave it out and the first planner takes that seat. Defaults are written by python3 scripts/llm_council.py configure to $XDG_CONFIG_HOME/llm-council/agents.json or ~/.config/llm-council/agents.json. Plans land as Markdown, run artifacts under ./llm-council/runs/<timestamp>.

Check the agents config before the first run: with no file present, the skill tells you to run ./setup.sh. Python 3 is required. Do not close the session while the council works — yielding before the 30-minute mark and before judge.md and final-plan.md are saved kills the run. Answering the intake questions is optional, but the skill notes that more detail improves the final plan.

What it helps with

  • get independent plans from several CLI agents for one task
  • remove one-model bias when picking an approach
  • merge competing plans into a single document with a judge's reasoning
  • survive a planner failure through retries
  • keep a run on disk for later review

Example prompts for the agent

  • run llm-council on this task and assemble the planner council
  • configure a council of Codex, Claude and OpenCode, then run it
  • run llm-council and save the final plan under llm-council/runs

How to install llm-council

1Skills CLI (Vercel)

npx skills add am-will/codex-skills --skill llm-council -a claude-code

2Manually: put the files in place

~/.claude/skills/llm-council/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/llm-council/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about llm-council

How do I install llm-council in Claude?

Run npx skills add am-will/codex-skills --skill llm-council -a claude-code. The skill lands at ~/.claude/skills/llm-council/SKILL.md.

What does llm-council do?

It launches several CLI agents as planners, anonymizes and shuffles their plans, then has a judge merge them into final-plan.md.

What does llm-council need to run?

Python 3 and a configured agents file; without one, the skill asks you to run ./setup.sh. The lineup is set in agents.planners.

Similar skills

All skills for Agent setup: memory, context, orchestration →

llm-council by other authors comes first.

llm-council aiwithremy llm-council sends a question or decision to five AI advisors, has them review one another, and produces a final verdict. Use it for high-stakes tradeoffs. CX Codex +1 1,400installs council boshu2 council collects separate judgments from several models or contexts and merges them into one report that keeps disagreements visible. CX Codex +1 3,000installs llm-context brave Extracts web page text, tables, and code chunks via Brave Search API for LLM grounding. Useful when building agentic RAG workflows. CX Codex +1 1,300installs jetson-llm-serve Jetson Team Creates a serving recipe for LLMs or VLMs on Jetson with vLLM or SGLang. Use it when an OpenAI-compatible model server is needed. CX Codex +1 1,200installs quark-torch-llm-ptq amd Walks a PyTorch or Hugging Face model through AMD Quark post-training quantization: intake, plan, manifest, run. CX Codex +1 524installs auto-review-loop-llm wanshuiyin Runs your work through review, fixes and another review until an outside model scores it high enough. Handy for pushing a paper or experiment to submission shape. CX Codex +1 496installs llm-eval-harness daymade This skill probes a third-party LLM endpoint across six dimensions: availability, request fidelity, speed, concurrency, protocol compliance and quality regression. CX Codex +1 372installs

From the same set10

Frontend Responsive Design Standards A skill that gives your coding agent rules for mobile-first layouts: breakpoints, relative units, touch targets and content priority. CX Codex +1 1,200installs role-creator Builds and updates Codex subagent definitions as standalone TOML files. Useful when you need isolated role configs with explicit model settings and developer instructions. CX Codex +1 1,200installs swarm-planner Builds implementation plans that show task dependencies for parallel agent work. Use it before coding changes in an unfamiliar codebase. CX Codex +1 1,200installs gemini-computer-use Runs Gemini 2.5 Computer Use agents with Playwright to drive browser workflows. It captures screenshots, calls UI actions, and prompts for confirmation on risky steps. CX Codex +1 1,200installs Agent Browser A skill that hands your agent a Rust-based headless browser: navigate, click, fill forms and snapshot pages through structured commands. CX Codex +1 1,200installs plan-harder Turns complex requests into structured sprint plans with atomic tasks. Triggers when you tell the agent to plan harder, focusing purely on architecture and risks. CX Codex +1 1,200installs read-github Lets the agent pull documentation and code from any GitHub repository through the gitmcp.io MCP service. Useful when you need to understand a project without cloning it. CX Codex +1 1,200installs planner Splits bug fixes and feature development into structured sprints and atomic tasks. It helps you design software changes before writing any code. CX Codex +1 1,200installs

Project sponsors

Companies that keep this analytics open
Gold sponsor
Nutra.Club

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor