# Coordinate Multi-Model Planning Councils

Complex coding changes often need more than one planning perspective. This skill runs several AI planner CLIs, compares their plans, and produces one consolidated implementation plan.

## Install

```bash
npx skillstore add am-will/llm-council
```

## Metadata

- - Slug: am-will-llm-council
- - Version: 1.0.0
- - Author: am-will
- - GitHub username: am-will
- - License: MIT
- - Repository: https://github.com/am-will/codex-skills/tree/main/skills/llm-council/
- - Ref: main
- - Supported tools: Claude, Codex, Claude Code
- - Risk level: high
- - Risk factors: scripts, external\_commands, network, filesystem, env\_access
- - Quality score: 38
- - Quality tier: warning
- - Public page: https://skillstore.pages.dev/skills/am-will-llm-council
- - Manifest: https://skillstore.pages.dev/api/skills/am-will-llm-council/manifest

## Capabilities

- Configures planner and judge agents for Codex, Claude, Gemini, OpenCode, or custom CLIs.
- Runs multiple planner agents and captures each Markdown plan as an artifact.
- Anonymizes and randomizes valid plans before the judge evaluates them.
- Produces judge output and a final Markdown implementation plan.
- Provides a local UI for monitoring, saving, accepting, and refining final plans.
- Stores run metadata, validation status, warnings, and plan files under a run directory.

## Use Cases

- Compare Architecture Plans: Generate independent implementation plans for a complex feature, then merge the strongest ideas into one plan.
- Review Risky Refactors: Ask several models to identify edge cases, test needs, and rollback steps before a large refactor starts.
- Standardize Planning Artifacts: Create consistent Markdown plans with validation, judge notes, and saved run history for team review.

## Prompt Templates

### Plan A Small Change

```
Use llm-council to plan this small code change: [describe change]. Ask me intake questions first, then run the council with default agents.
```

### Compare Refactor Approaches

```
Use llm-council to compare refactor plans for [module]. Include constraints, affected paths, test expectations, and rollback requirements in the task brief.
```

### Audit A Migration Plan

```
Use llm-council to plan a migration from [current system] to [target system]. Require each planner to cover compatibility, data safety, deployment, monitoring, and rollback.
```

### Run A Custom Council

```
Use llm-council with my configured agents to produce a final plan for [initiative]. Keep custom command use limited to the trusted config I provide.
```

## Limitations

- Requires the selected AI CLIs to be installed and authenticated before use.
- May send task and repository context to external model providers through those CLIs.
- Plan quality depends on the configured models and the task brief quality.
- Custom commands can execute local programs and should be used only with trusted configs.

## Best Practices

- Use a clear task brief with constraints, affected paths, success criteria, and required tests.
- Review generated plans before implementation because planner and judge outputs are untrusted content.
- Run custom commands only from trusted configuration files in repositories that can safely use external AI tools.

## Anti Patterns

- Do not use this skill on confidential repositories unless provider data policies allow the selected CLIs.
- Do not paste secrets, credentials, or production tokens into task briefs or refinement requests.
- Do not configure arbitrary custom commands from untrusted users or copied examples.

## Security Audit

- - Safe to publish: false
- - Audited at: 2026-06-28T08:29:40.992\+00:00
- - Summary: Static analysis produced many hits, but most Markdown backtick, weak-crypto, path traversal, and C2 keyword detections are false positives. The confirmed risks are high-impact: the skill launches external CLI agents, supports custom commands, uses a permission-bypass CLI flag, writes artifacts, and exposes local UI state. No evidence found of malicious credential theft or command-and-control intent, so this is high risk rather than blocked critical.

## Stats

- - Views: 177
- - Downloads: 10
- - Favorites: 2
- - Popularity score: 0