The Agent Skills Directory

[SAFE]: The skill is a legitimate authoring toolkit designed to facilitate the creation of other skills. It contains no evidence of malicious patterns, obfuscation, or unauthorized data access. It explicitly documents and enforces Anthropic's safety guidelines for skills, such as prohibiting network access and runtime package installations within the generated skills.\n- [COMMAND_EXECUTION]: The skill includes and utilizes a local Python script (scripts/validate_skill.py) intended for validating the structure and content of generated skills. The script uses standard Python libraries (argparse, pathlib, re) to perform local filesystem checks and does not perform network operations. Examples also include templates for automation scripts (e.g., scripts/analyze.py) which are provided as placeholders and use only standard library components.\n- [INDIRECT_PROMPT_INJECTION]: As a tool that generates instructions based on user input, the skill possesses an inherent surface for indirect prompt injection where a user could potentially supply malicious instructions to be included in a skill. However, the skill incorporates mitigation measures such as structured templates and a validation tool that scans for reserved words, prohibited content (like XML tags), and improper formatting to maintain the integrity of the generated output.

creating-skills