scene-structure-techniques
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill framework ingests user-supplied content to generate and store scene outlines, which presents a surface for indirect prompt injection attacks.\n
- Ingestion points: Interactive prompts for scene goals, obstacles, and rhythmic beats found in the template placeholders of SKILL.md.\n
- Boundary markers: The template lacks specific delimiters or instructions to treat user-provided content as non-executable data.\n
- Capability inventory: The skill references file system write operations for saving outlines and utilizes the Read tool.\n
- Sanitization: No input validation or content sanitization processes are defined for the user-supplied text.
Audit Metadata