skill-creator
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided inputs such as skill concepts, use cases, and capability definitions to generate documentation and skill templates. This creates a surface where malicious instructions in the input could influence the generated output. \n
- Ingestion points: User-provided data in Phase 1 (Conception) and Phase 2 (Design) worksheets in SKILL.md.\n
- Boundary markers: Absent; the skill uses fill-in-the-blank placeholders but does not implement explicit delimiters or safety instructions for handling external data.\n
- Capability inventory: File writing to the .skill-sandbox/ directory (Phase 3.1) and command execution via Node.js (Phase 3.2).\n
- Sanitization: None observed; inputs are directly interpolated into the generation process.\n- [DYNAMIC_EXECUTION]: The skill facilitates the creation of new executable artifacts (SKILL.md files) and instructs the agent to run validation scripts against them at runtime. This allows for the generation and immediate testing of code based on user definitions.\n- [COMMAND_EXECUTION]: The skill directs the agent to execute local shell commands using node to run validation and evaluation scripts (e.g., node .skill-sandbox/validations/validate-sandbox-skills.mjs) found within the environment's platform and sandbox directories.
Audit Metadata