manim-composer
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill creates a potential surface for indirect prompt injection by processing untrusted data from web research and user concepts to generate scene plans.\n
- Ingestion points: Topic research in Phase 1 (web search) and clarification questions in Phase 2 (user input).\n
- Boundary markers: The skill does not define boundary markers or explicit instructions to delimit and ignore potentially malicious content within the external research data.\n
- Capability inventory: The skill possesses web search capabilities and the ability to generate markdown files (scenes.md).\n
- Sanitization: The instructions do not specify any sanitization, filtering, or validation of content retrieved during the research phase.\n
- Downstream impact: Maliciously crafted content from web searches could influence the generated scene plans, potentially affecting subsequent implementation steps when other skills or agents process the generated scenes.md file.
Audit Metadata