What problem does it solve? When multiple AI coding agents (Codex, Cursor, Cline, Copilot, Gemini CLI, etc.) share a home directory, they recursively load each other's SKILL.md and AGENTS.md files, causing silent tool-name mismatches and wrong behavior. This Skill provides a governance pattern that isolates each agent while keeping skills shareable. ## Core Features & Use Cases - Three-Plane Governance Model: Separates skills into a canonical source-of-truth plane, per-agent runtime planes, and a librarian documentation plane with clear edit rules. - Adaptation Pipeline: Translates tool names, paths, and conventions (e.g., Task → delegate_task, Bash → terminal) when porting skills between platforms, with a Python adaptation script template. - Contamination Auditing: Provides shell commands to inventory every SKILL.md and AGENTS.md on disk, detect foreign-directory loads, and verify runtime scoping. - Use Case: You install Cline alongside Codex and notice Cline referencing tools it doesn't have. Use this Skill to scope Cline to ~/.cline/skills/, adapt your canonical skills with correct tool names, and build an ecosystem map to track drift. ## Quick Start Audit my home directory for all SKILL.md files and help me set up the three-plane model so each of my AI coding agents only loads skills from its own directory.
