The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from user requests and transcripts to populate the text content of generated video scenes.
Ingestion points: Step 2 instructions specify using the user's words or speaker's words from a transcript.
Boundary markers: The instructions do not define specific delimiters or warnings to ignore embedded instructions within the ingested text.
Capability inventory: The skill uses ffmpeg for frame extraction, get_clip_frame for verification, and duplicate_motion_document for document management.
Sanitization: No explicit sanitization or escaping of the ingested text is mentioned.
Contextual note: The risk is limited as the text is intended for rendering as visual nodes within a motion design document rather than for execution in a shell or prompt interpolation.
[COMMAND_EXECUTION]: The skill uses the ffmpeg command-line tool to extract frames from video examples for visual analysis.
Evidence: The instruction in Step 1 provides a specific command for generating a tiled contact sheet from a video file: ffmpeg -i gallery/network.mp4 -vf fps=2,scale=480:-1,tile=4x4 sheet.png.
Note: This is a standard utility usage for a motion design workflow and targets internal gallery files rather than user-supplied paths.
[EXTERNAL_DOWNLOADS]: The skill requires loading external design modules managed by the vendor.
Evidence: Step 3 instructs the agent to load tella-motion-design, tella-title-cards, and tella-explainer-graphics.
Note: These resources are authored by the skill's vendor (tellahq) and represent standard modularized design rules.

grainy-gradient-scenes