enforce-safe-skills

What problem does it solve?

Agents working with shared skill libraries, protected source repositories, or reusable tool folders often accidentally modify canonical source files, push unapproved changes, or add agent-specific metadata to shared repos, breaking the library for all users and violating authorization rules.

Core Features & Use Cases

• Copy-First Adaptation Enforcement: Requires agents to copy needed skills to their local environment instead of modifying the protected source repo directly.
• Explicit Approval Requirement: Blocks all source repo mutations (including reformatting, adding metadata, or pushing commits) unless the user provides clear, specific approval for the exact change.
• Secret and Customization Guardrails: Prevents importing secrets into shared repos and ensures agent-specific customizations are confined to local agent environments.
• Use Case: If your team maintains a shared library of AI skills, this skill ensures you only make local changes to your own copy and get explicit approval before altering the shared source, keeping the library stable and consistent for all team members.

Quick Start

Use the enforce-safe-skills skill to copy the required design skill from the shared protected repository to your local environment and make any needed adaptations without modifying the original source.