The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from component source code, stories, and call sites which serves as an attack surface.
Ingestion points: Component code, props types, styles, and stories are read in Step 2, and call sites are retrieved in Step 3.
Boundary markers: There are no instructions or delimiters provided to prevent the agent from executing instructions that might be embedded within the source code or stories.
Capability inventory: The skill writes documentation files and example files to the filesystem and utilizes the grep tool for analysis.
Sanitization: The instructions lack sanitization requirements; Step 5 explicitly commands the agent to copy example code "character for character" into the generated documentation.
[COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command (grep -o) on a file path derived from the component name (<name>). This pattern creates a potential vulnerability for path traversal or command injection if the component name is supplied by a user and contains malicious characters.

component-docs