process-transformation-story
Pass
Audited by Gen Agent Trust Hub on Aug 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a structured 'reference contract' (Preserve/Replace/Avoid) which encourages the agent to selectively use logic from examples rather than blindly copying content, serving as a functional security boundary.
- [SAFE]: The 'Safety and attribution' section explicitly prohibits the agent from inventing original sources, making unauthorized account changes, or purchasing assets, which aligns with security best practices for agentic tools.
- [INDIRECT_PROMPT_INJECTION]: The skill establishes a data ingestion surface by requiring the agent to process external prompt excerpts and media content.
- Ingestion points: Content is read from the
references/cases.mdfile and external media assets are accessed via the browser or specialized image/video tools. - Boundary markers: The agent is instructed to create a 'reference contract' to explicitly define what to preserve or avoid, and is warned not to impersonate creators or claim visual fidelity if media cannot be verified.
- Capability inventory: The skill utilizes the agent's ability to read local files, access the internet via a browser tool, and process visual media via image/video tools.
- Sanitization: The workflow requires a comparison between the generated artifact and the anchor case, with a mandatory revision step to correct failures, serving as a qualitative check on the output.
Audit Metadata