web-sims
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell commands using Node.js, Python 3, and ffmpeg (e.g.,
node sims/build.mjs,python3 sims/<name>/sound.py,ffmpeg) which are located in an external repository.- [EXTERNAL_DOWNLOADS]: The skill metadata and documentation reference an external GitHub repository (https://github.com/whaleyxbt/claude-motion) as a prerequisite, introducing a dependency on external code that the agent is expected to execute.- [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface that ingests potentially untrusted data through project files and user input. - Ingestion points: Ingests simulation beats and configuration from
timeline.jsonand user descriptions provided in the agent's context. - Boundary markers: No explicit markers or warnings are defined to distinguish between data and instructions.
- Capability inventory: Access to shell execution for Node.js, Python, and ffmpeg is specified in
SKILL.md. - Sanitization: No validation or sanitization logic is described for the input data before it influences script execution.
Audit Metadata