The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md file utilizes the !command syntax to load project-specific protocols and configurations. The commands, such as cat .production-grade.yaml and reading from skills/_shared/protocols/, are used for environment setup and do not involve data exfiltration or sensitive system file access.
[COMMAND_EXECUTION]: The LITE.md file contains instructions for the agent to execute shell commands to "ground" itself in the project. These include inspecting package.json for styling frameworks (Tailwind, Framer Motion), searching the docs/ directory for design assets, and checking for visual regression tests. These are standard developer workflow actions.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow where external data (UX research, visual basis documents) is ingested to produce code and technical specifications, creating a potential surface for indirect injection.
Ingestion points: UX Research findings, Visual Basis documentation, and project-specific user authority inputs.
Boundary markers: The skill does not define explicit delimiters to isolate external research data from the core instructions.
Capability inventory: The skill generates production-grade Framer Motion (TypeScript) components, CSS animations, and markdown handoff specifications.
Sanitization: No specific sanitization or filtering logic is provided for the data processed during the translation from research to specification.

interaction-designer