create-composition

Pass

Audited by Gen Agent Trust Hub on Sep 30, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The documentation provides standard boilerplate for web-based animation. All referenced functionality is restricted to the browser context (Canvas API and DOM) and the specific animation framework, representing standard usage.
  • [INDIRECT_PROMPT_INJECTION]: The skill identifies a pattern for ingesting data to render visual content. 1. Ingestion points: The helios.subscribe function in SKILL.md receives state.inputProps which can contain user-provided strings. 2. Boundary markers: A JSON schema is defined to validate input properties, providing a structured boundary. 3. Capability inventory: The skill only utilizes canvas rendering (ctx.fillText) and basic DOM manipulation; it has no file system, network, or subprocess access. 4. Sanitization: Input is validated against the schema before rendering.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 30, 2026, 01:44 PM