Skills
skills/davekilleen/dex/anthropic-skill-creator/Gen Agent Trust Hub

anthropic-skill-creator

Warn

Audited by Gen Agent Trust Hub on Mar 7, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/init_skill.py performs directory creation and file writes based on user-supplied paths and skill names.
  • [COMMAND_EXECUTION]: The initialization script scripts/init_skill.py uses chmod(0o755) to set executable permissions on a dynamically created Python script (scripts/example.py).
  • [COMMAND_EXECUTION]: The initialization script scripts/init_skill.py lacks path validation for the user-provided skill name and path, which could potentially allow path traversal during the creation of the skill directory.
  • [COMMAND_EXECUTION]: The script scripts/package_skill.py reads from the local filesystem and creates a compressed .skill archive (ZIP format) using the zipfile module.
  • [PROMPT_INJECTION]: The skill serves as a template and structure generator, creating a surface for indirect prompt injection where untrusted data provided by a user during skill creation could be incorporated into instructions or scripts processed by an agent.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 7, 2026, 01:48 PM