The Agent Skills Directory

[SAFE]: The skill serves as a pedagogical resource for developing other skills. It defines clear standards for naming conventions, directory structures (scripts, references, assets), and progressive disclosure to optimize context usage.
[SAFE]: The environment-reference.md file provides a detailed and accurate catalog of pre-installed Python packages, Node.js tools, and system utilities available in the agent's execution environment. This transparency helps prevent unnecessary exploratory command execution.
[SAFE]: Instructions for packaging skills using standard utilities like zip and delivering them via the computer:/// protocol follow established platform safety patterns for file handling.
[SAFE]: The skill provides benign templates for validation and transformation scripts, encouraging the use of deterministic code for complex operations while emphasizing error handling and exit codes.
[SAFE]: Network activity and package installation instructions are limited to well-known, trusted registries (PyPI, NPM, GitHub) and official platform domains, adhering to standard development practices.

creating-skill