The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-supplied data such as shot scripts, character descriptions, and creative briefs. This content is then interpolated into prompts used by the model_run tool to generate video assets. This workflow creates an attack surface where instructions embedded in the user data could potentially influence or override the agent's intended operation.\n
Ingestion points: Identified in SKILL.md (Step 1: Script) and references/board-craft.md (Drawing the panels).\n
Capability inventory: The skill uses numerous tools including model_run, search, upload_asset, asset_get, and collection_create throughout its operation.\n
Boundary markers: The skill employs structured prompt templates in references/video-prompt.md (e.g., 'TIMELINE', 'REFERENCE ROLES'), which serve as organizational delimiters.\n
Sanitization: The skill lacks explicit sanitization or filtering mechanisms for the user-generated content before it is included in model prompts.\n- [COMMAND_EXECUTION]: In SKILL.md, the agent is instructed to suggest the command npx skills add scenario-labs/skills to the user if sibling skills are missing. This command facilitates the installation of external components from the vendor's repository.

scenario-seedance-storyboard