product-hero-scene-planner

Pass

Audited by Gen Agent Trust Hub on Nov 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input in the form of user prompts and upstream briefs to generate scene plans.
  • Ingestion points: Defined in SKILL.md under Required Inputs and contract.json (user prompt, upstream brief, and get_objects_summary data).
  • Boundary markers: The skill instructs the agent to restate intent in plain language and maintain an assumptions register to document unverified inputs.
  • Capability inventory: The skill is restricted to generating text plans. It explicitly forbids executing Python code, Blender CLI commands, or producing artifacts.
  • Sanitization: No automated sanitization is described, but the agent is limited to producing structured text output without code execution.
  • [SAFE]: The skill includes informational links to well-known technology documentation sites, including the official Blender Foundation and Anthropic Claude resources. These links are for reference and do not involve the download of executable scripts or packages.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 15, 2026, 01:08 AM