runway-reference-architecture
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill instructions promote security best practices, including isolating credentials from the client side, defining trust zones, and ensuring secret access is restricted to server-side workers.
- [SAFE]: The 'Output' section explicitly forbids the exposure of API secrets, prompt contents, media contents, or temporary signed URLs in the agent's responses.
- [SAFE]: External resource links point to the official RunwayML documentation (docs.dev.runwayml.com), which is a well-known service domain.
- [SAFE]: The 'Tool Discipline' section provides clear constraints, instructing the agent to use inspection tools (Read, Grep) for analysis and limiting modification tools (Write, Edit) to approved changes only, while explicitly prohibiting unauthorized production mutations.
Audit Metadata