The Agent Skills Directory

[SAFE]: No malicious instructions, prompt injection patterns (like DAN or ignore commands), or obfuscated content were found in the provided file. Its behavior is limited to project data management within specified directories.\n- [INDIRECT_PROMPT_INJECTION]: The skill includes instructions to process external data sources which is essential for its primary purpose. Analysis identifies the following surface: 1. Ingestion points: Reads standalone contracts, screenplay scenes, and director books (SKILL.md). 2. Boundary markers: No explicit delimiters are specified for external content. 3. Capability inventory: Instructions include writing structured output to '03_preproduction/previz/SCN-###.json' and 'SCN-###.md' (SKILL.md). 4. Sanitization: No sanitization steps are defined for input data. However, the risk is categorized as safe because it is tied to the primary function and the skill lacks network or execution capabilities.

previz-plan