The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill mentions the installation of sibling skills from the vendor's repository (scenario-labs/skills) via npx. This is a standard recommendation for users to extend functionality within the same vendor ecosystem.
[INDIRECT_PROMPT_INJECTION]: The skill processes image assets and text prompts, creating a surface for potential injection from external content. However, the instructions focus on technical parameter management and prompt precision for image generation, with no evidence of malicious intent or behavior.
Ingestion points: Image assets (IDs) and user-supplied prompts processed by image models in SKILL.md.
Boundary markers: Instructs the agent to use specific parameter structures and quote strings exactly in prompts to ensure fidelity.
Capability inventory: Uses tools for model_run, asset_display, and asset_download.
Sanitization: Relies on the platform's tool schemas and job management lifecycle to handle inputs.

scenario-seedream