The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied data such as scripts and images to generate video content. This constitutes a standard data ingestion surface.
Ingestion points: Processes user-supplied scripts, recorded narration, portrait photos, and product images (SKILL.md).
Boundary markers: The instructions explicitly mandate that the agent only use lines supplied or approved by the user and avoid fabricating claims.
Capability inventory: Uses tools like model_run, upload_asset, and model_scenario-compose-video across the workflow (SKILL.md).
Sanitization: Includes a mandatory verification step (scenario-asset-analysis) to audit generated frames against original uploads to detect hallucinations or label drift.
[COMMAND_EXECUTION]: The skill instructs the agent to ask the user to run npx skills add scenario-labs/skills if required sibling skills are missing. This command refers to the official repository of the skill author (scenario-labs) and is a standard procedure for skill management within this ecosystem.

scenario-ugc