The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided brand definitions to generate configuration files and design tokens on the local filesystem.
Ingestion points: User input defining brand positioning, colors, and typography during the definition and audit steps in SKILL.md.
Boundary markers: The agent is guided by a structured 10-step workflow, and mathematical calculations are offloaded to a dedicated script (scripts/tokens.py) for deterministic results.
Capability inventory: The skill involves writing a brand.json configuration file and executing a Python tool that generates tokens.css and tokens.json files in an output directory.
Sanitization: The scripts/tokens.py script includes validation logic to ensure color values are valid hex codes and that scale parameters are positive numbers, mitigating basic malformed data risks.

brandkit