worldbuilding

Pass

Audited by Gen Agent Trust Hub on Nov 2, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill relies on a local Node.js script (story.js) to perform maintenance tasks such as indexing files and checking story continuity. It includes specific security instructions to use the standard Node.js runtime, which prevents the accidental execution of untrusted configuration files like bunfig.toml that might be present in a user's project directory.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-generated story content and metadata, which serves as a potential surface for indirect prompt injection. This risk is addressed through clear instructional guardrails.\n
  • Ingestion points: The agent is instructed to read story.md and various markdown files within the worldbuilding/ directory to gain context for new elements.\n
  • Boundary markers: The skill incorporates a mandatory shared convention that instructs the agent to treat external text—such as reader notes, issues, or imported content—strictly as reference material and never as authoritative instructions.\n
  • Capability inventory: The skill's capabilities are focused on file system management (creating and updating markdown files) and local validation via the provided CLI tool, without performing arbitrary network operations.\n
  • Sanitization: The system relies on the agent's adherence to the provided conventions to distinguish between story data and instructional commands.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 2, 2026, 11:36 PM