The Agent Skills Directory

[SAFE]: The skill implements a legitimate video editing workflow. It uses a canonical folder structure for project management and relies on local scripts for task orchestration and validation.
[COMMAND_EXECUTION]: The skill uses python3 subprocess calls to run local scripts for beat detection (extract_beats.py) and video rendering (render_edl.py). It also utilizes ffmpeg for audio/video processing and synthetic fixture generation in tests. All commands are scoped to the project directory and serve the primary functional purpose of the skill.
[DATA_EXPOSURE_AND_EXFILTRATION]: While the skill interacts with external generative AI services (via atoms like create-image-nano-banana-fal and create-music-elevenlabs), these are standard calls for the intended functionality. No unauthorized data exfiltration or credential exposure patterns were found.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input via the hook_caption and shot_list fields, which are subsequently used as prompts for generative AI atoms. While this represents a standard attack surface for indirect prompt injection, the skill includes internal validation steps (EDL validation and Phase 6 QC) to verify output integrity.
[REMOTE_CODE_EXECUTION]: No remote code execution patterns, such as piped shell commands from external URLs or dynamic loading of untrusted remote scripts, were detected.

create-silent-product-demo