The Agent Skills Directory

[SAFE]: The skill provides structured guidance and helper scripts for a developer workflow. Analysis of the included files shows no malicious patterns.
[COMMAND_EXECUTION]: The skill instructions and scripts (discover_skills.py, scaffold_skill.py, validate_skill.py) involve standard filesystem operations and local process execution using the Python subprocess (via shell examples in SKILL.md) and pathlib modules. These are used for project management tasks (creating directories, reading/writing files, searching local folders) and do not involve network operations, privilege escalation, or persistence mechanisms.
[INDIRECT_PROMPT_INJECTION]: The skill consumes user-provided strings for skill names and descriptions during the discovery and scaffolding processes. However, the scripts perform normalization and validation (e.g., normalize_skill_name in scaffold_skill.py and frontmatter parsing in validate_skill.py), and the resulting data is used to generate local documentation rather than being executed as instructions. The risk is assessed as safe given the development-oriented purpose and lack of sensitive capability tiers.
[DATA_EXPOSURE]: The scripts read local files within a specified --skills-dir. This behavior is transparent, documented, and restricted to the directory provided by the user. No credentials or sensitive system paths are accessed.

skill-creator