create-skill

Pass

Audited by Gen Agent Trust Hub on Feb 25, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The task skill template includes instructions to install the 'uv' tool from its official domain (astral.sh). This targets a well-known technology provider and is documented neutrally as a safe prerequisite.
  • [REMOTE_CODE_EXECUTION]: The skill facilitates the creation of task-based skills that execute Python scripts via 'uv run'. It also provides templates recommending dynamic module loading via 'sys.path.insert' for internal skill integration, which is part of the architectural design.
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection as it constructs instructional files (skills) based on untrusted user requirements.
  • Ingestion points: User requirements for skill purpose and logic are gathered in Step 1 of SKILL.md.
  • Boundary markers: No specific delimiters are used to isolate user-provided text within the generated SKILL.md files.
  • Capability inventory: The skill possesses the capability to write files to the local filesystem during the creation process (Step 9).
  • Sanitization: User input is interpolated into templates without validation or escaping, allowing potentially malicious instructions to be persisted in newly created skills.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 25, 2026, 09:10 PM