The Agent Skills Directory

[SAFE]: The skill provides standard documentation for extending a frontend framework. It uses Zod for data validation and React for component rendering.
[COMMAND_EXECUTION]: The skill instructs the user to run a local codegen script (pnpm gen:embed-widgets) and linting/typechecking commands. These are standard development tasks for maintaining consistency between the frontend and backend schemas.
[INDIRECT_PROMPT_INJECTION]: The skill documents an attack surface where descriptions and examples provided in the schema are passed directly to an LLM's system prompt. However, it provides clear guidance on how to define these schemas safely and includes validation steps (Zod, manual testing) to mitigate risks. This is a functional requirement of the feature described.

seer-embed