stop-that-shit

Overview of stop-that-shit

What it is

Stop That Shit (别再造史了) is a multi-platform Hook + Skill Guard for AI coding agents in Codex/GPT workflows. It intercepts unrequested hashes, checksums, and task-scope creep. The project is licensed under MIT and categorized as Developer Tools. It supports Codex, Claude Code, OpenCode, Hermes Agent CLI, and Pi Coding Agent.

Key features

• Defines four types of unwanted agent behavior: Scope creep, Hashing & hypothetical hardening, Intent violation, and Task thrashing (SHIT).
• Provides a Skill for engineering judgment and a Guard for checking explicit boundaries before supported tool calls.
• Includes a set of 18 public Bad/Good test cases for validating the rules.
• Offers a separate companion tool called Stop That Shit Slop (STSS) for removing unnecessary defensive text.
• Supports multiple deployment modes: as a plugin, a global Skill, or a separate extension.

Use cases

• Preventing an AI coding agent from generating SHA-256 checksums or other hashes that are never read by anyone.
• Stopping an agent from refactoring unrelated modules when it was only asked to review or fix a specific issue.
• Preventing an agent from repeatedly re-checking work that has already been verified and has no new issues or changes.
• Ensuring an agent does not modify files when instructed to only read and report.
• Cleaning up unnecessary defensive or hedging language in generated text or documentation.

Getting started

See the repository README for full installation instructions for each host.For example with Claude Code plugin (requires Node.js 18+):bashclaude plugin validate .claude plugin marketplace add ./claude plugin install stop-that-shit@stop-that-shitThen restart Claude Code or run /reload-plugins, and use commands like /stop-that-shit:stop-that-shit review -- ....

License and pricing

License: MIT. Pricing: open source. The project is free to use. The license is stated as MIT in the metadata.

Based on the project's public README and repository metadata.