playful-lab-3d
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill transforms untrusted user content like outlines, articles, or product info into a 3D HTML deck. (1) Ingestion points: User-supplied data is embedded into the PAGES configuration array within the generated deck.html. (2) Boundary markers: The template lacks explicit delimiters or instructions to the agent to ignore malicious commands embedded in the user text. (3) Capability inventory: The output is a functional web page using Three.js and triggering a local file-writing Python script for export. (4) Sanitization: There is no guidance provided for escaping or validating user content before it is interpolated into the HTML context.
- [DYNAMIC_EXECUTION]: The skill instructions involve programmatically modifying a JavaScript-heavy HTML template (PAGES array) to generate the final deck, which constitutes simple script generation from a template.
- [COMMAND_EXECUTION]: The README and SKILL files provide instructions for executing shell commands to run a local HTTP server and a specific Python export script located in the scripts directory.
Audit Metadata