The Agent Skills Directory

[COMMAND_EXECUTION]: The provided scripts (new-skill.sh, lint-skill.sh, lint-frontmatter.sh) utilize standard system utilities such as mkdir, cp, sed, ln, awk, and rm. These operations are strictly for managing the skill lifecycle, including scaffolding directories and creating symlinks for discovery in agent-specific configuration paths like ~/.agents/skills.- [DYNAMIC_EXECUTION]: The lint-frontmatter.sh script dynamically executes Python or Node.js code snippets to parse YAML frontmatter from SKILL.md files. This is a local operation used to validate that skill metadata is correctly formatted and compatible with various agent platforms.- [INDIRECT_PROMPT_INJECTION]: The skill processes skill names, descriptions, and existing SKILL.md content during scaffolding and linting. While this constitutes a surface for processing user-generated instructions, the scripts perform structural analysis (e.g., checking line counts, regex matching for shebangs, and YAML parsing) rather than interpreting or executing the instructions. The scaffolding script includes input validation for skill names.- [SAFE]: The skill is a well-documented developer utility that facilitates the creation of standardized, high-quality agent skills. It explicitly promotes security best practices, such as pinning versions for external tool invocations and providing dry-run functionality for destructive operations.

skill-author