The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied brand colors and token arguments to generate palettes and modify CSS/JSON files. This creates a surface for indirect prompt injection if malicious instructions are embedded in the provided data.
Ingestion points: User-provided brand colors and token arguments in the generate, add, and update commands.
Boundary markers: The instructions do not define explicit delimiters to isolate external data from the agent's prompt context.
Capability inventory: The skill utilizes Write, Edit, and Bash tools to manage project files in the .styleseed, tokens, and css directories.
Sanitization: Validation is present in scripts/generate-palette.mjs, which enforces hex color formats and checks allowed value sets for design parameters.

ss-tokens