The Agent Skills Directory

[SAFE]: The skill defines a legitimate and structured post-production workflow. It utilizes local CLI tools (cinefoundry, ffmpeg) and standard AI models (Qwen3-TTS, Whisper) for audio-visual tasks without requesting excessive permissions or performing suspicious network operations.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection through the ingestion of user-provided script data.
Ingestion points: User-provided script fields text (for titles) and vo (for voice-over/subtitles) are processed in Workflow steps 4 and 5 in SKILL.md.
Boundary markers: Not explicitly mentioned in the instructions for the agent.
Capability inventory: The skill uses cinefoundry CLI and ffmpeg for rendering assets and delivery.md file writing.
Sanitization: The skill includes a specific check for font glyph compatibility (Noto Sans CJK SC) before rendering, which helps prevent rendering-based attacks. The risk is assessed as low because the injected text is primarily used for static rendering rather than command execution or logical branching.

final-cut-assembly