agents

This repository provides a standardized system for managing AI agent instructions via a canonical `AGENTS.md` file. It includes tools to synchronize these instructions across various AI tools (such as Claude Code, Cursor, Gemini, and Copilot) and implements a comprehensive suite of safety and quality checkers. Key capabilities include: - **Instruction Synchronization**: Scripts to maintain consistent rule sets across multiple tool-specific configuration files (e.g., `.cursorrules`, `.clinerules`, `CLAUDE.md`). - **Safety Guardrails**: A collection of portable Python checkers to prevent destructive bash commands, infrastructure access, and the use of banned agents. - **Compliance Tooling**: Automated checks for git identity, branch naming, secret heuristics, and weak hashing algorithms. - **Prose Analysis**: Advisory tools to analyze commit messages, pull request titles, and general documentation for hedging, passive voice, and banned vocabulary. - **Workflow Integration**: GitHub Actions workflows for enforcing compliance on pull requests and pushes, including immutable-tree orchestration. - **Agent Hooks**: Specialized hooks for Claude Code to enforce consent gates and block high-risk operations.