The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/validate_skill.py script fetches the skills-ref package from a remote Git repository (github.com/agentskills/agentskills.git) using the uv package manager.\n- [REMOTE_CODE_EXECUTION]: The validation process relies on executing code downloaded directly from a GitHub repository via uv run, which introduces a dependency on the integrity of the remote source.\n- [COMMAND_EXECUTION]: The initialization script scripts/init_skill.py uses chmod(0o755) to set executable permissions on the newly created scripts/example.py file.\n- [COMMAND_EXECUTION]: The skill frontmatter suggests enabling broad tool permissions, specifically for Bash(python3:*) and Bash(uv:*), which allows the agent to execute Python scripts and manage packages.\n- [PROMPT_INJECTION]: The skill provides an interface for generating new system instructions from user-provided input, creating a surface for indirect prompt injection.\n
Ingestion points: User input provided to define the name and description fields in the generated SKILL.md file.\n
Boundary markers: None; the skill directly interpolates user input into the template file without instruction wrappers or safety markers.\n
Capability inventory: The skill utilizes Write, Edit, and Bash tools to manage and execute content on the local filesystem.\n
Sanitization: No input validation or character escaping is performed on the user-provided strings before they are written to disk.

skill-creator