The Agent Skills Directory

[SAFE]: The skill is entirely instructional and does not contain any malicious scripts or executables. All analyzed content is focused on filmmaking methodologies and prompt engineering guidelines.
[PROMPT_INJECTION]: While the skill establishes a "Director" persona to guide model responses, it does not attempt to override the AI's core safety filters, bypass system boundaries, or extract underlying instructions. The persona is used exclusively for artistic alignment.
[DATA_EXPOSURE]: No hardcoded API keys, tokens, or sensitive credentials were found. File path references (e.g., to .env or .ssh) are absent. Placeholders like [图URL] are used legitimately to demonstrate image-to-image prompting techniques.
[OBFUSCATION]: There is no hidden or encoded content. The hexadecimal codes present across the aesthetic files are valid color definitions used for cinematography color grading.
[REMOTE_CODE_EXECUTION]: The skill does not download or execute remote scripts. References to external platforms like Midjourney, Seedance, and Flux are descriptive and relate to the formatting of prompts for those services.
[DYNAMIC_CONTEXT_INJECTION]: The main skill entry point (SKILL.md) contains only metadata and instructions; it does not use the shell execution syntax (!command) to perform silent or automated shell operations.

cinema-director