By jarfa
Start a dialogue between two subagents to critique & improve your plans or code
npx claudepluginhub jarfa/critique-loop --plugin critique-loopA Claude Code plugin that starts a dialogue between two subagents to critique & improve your plans or code.
When planning or writing code I often want Claude Code to critique its own work. Previously this required manually running two instances and relaying messages between them.
This plugin automates the process: invoke once, and a two subagents are spawned to provide critique, review, or collaboration. The human stays out of the loop until the dialogue reaches a conclusion.
Use this for:
From GitHub:
# Add the marketplace
/plugin marketplace add jarfa/critique-loop
# Install the plugin
/plugin install critique-loop@jarfa
From local source (for development):
claude --plugin-dir /path/to/critique-loop
Just describe what you want in plain text:
/critique-loop plan my new caching strategy
/critique-loop review the auth module with codex
The plugin infers appropriate roles (e.g., "proposer" / "critic" for planning, "author" / "reviewer" for reviews) and starts the dialogue automatically. Say something like "with codex" in your description to use Codex as the second partner instead of Claude.
| Environment Variable | Default | Purpose |
|---|---|---|
DIALOGUE_DIR | .dialogues | Where dialogue files are stored (project-local) |
The dialogue file currently serves as the communication medium between the subagents. A potential optimization: direct communication where the orchestrator passes context in the resume prompt and the partner returns its response directly, eliminating file I/O overhead. The dialogue file would then become optional for debugging purposes.
This project is licensed under the MIT License — see the LICENSE file for details.
Access thousands of AI prompts and skills directly in your AI coding assistant. Search prompts, discover skills, save your own, and improve prompts with AI.
Behavioral guidelines to reduce common LLM coding mistakes, derived from Andrej Karpathy's observations on LLM coding pitfalls
Design fluency for frontend development. 1 skill with 23 commands (/impeccable polish, /impeccable audit, /impeccable critique, etc.) and curated anti-pattern detection.
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Reliable automation, in-depth debugging, and performance analysis in Chrome using Chrome DevTools and Puppeteer
Claude Code skills for Godot 4.x game development - GDScript patterns, interactive MCP workflows, scene design, and shaders