skill-creator

Pass

Audited by Gen Agent Trust Hub on Mar 2, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes local Python scripts (init_skill.py, package_skill.py) to perform file system operations, such as creating directories and writing files, which is consistent with its role as a project initializer.
  • [COMMAND_EXECUTION]: In init_skill.py, the script programmatically adjusts file permissions (chmod 0o755) for generated example files. This facilitates the creation of executable scripts but represents a deliberate modification of file security settings.
  • [PROMPT_INJECTION]: The creator command presents an indirect prompt injection surface by taking user-supplied descriptions and incorporating them into new SKILL.md and plugin.json files.
  • Ingestion points: User input provided to the /skill-creator:creator command in commands/creator.md.
  • Boundary markers: The generation process does not employ explicit delimiters or system instructions to prevent the user input from being interpreted as instructions by an agent loading the new skill.
  • Capability inventory: The skill has the capability to write to the file system and create executable files through the init_skill.py utility.
  • Sanitization: Validation logic in quick_validate.py checks for specific characters like angle brackets and validates naming conventions, but it does not provide comprehensive protection against malicious instructions embedded in the natural language description.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 2, 2026, 03:55 PM