The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions (Workflow Step 1) direct the agent to read and process an external file named DESIGN.md to extract design tokens for typography, color, and layout.
Ingestion points: The workflow in SKILL.md explicitly requires reading DESIGN.md as the primary source for design tokens.
Boundary markers: There are no specified delimiters or instructions to ignore potential commands embedded within the external design file.
Capability inventory: The skill is configured with fs:write and prompt:inject capabilities in open-design.json, which represent the available attack surface if the ingested data contained malicious instructions.
Sanitization: No explicit sanitization, validation, or escaping of the design token values is defined in the instructions.

motion-frames