The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest user prompts describing complex logic and UI requirements to generate functional prototypes. This creates a vulnerability surface where instructions embedded in the user's data could be reflected into the generated code.
Ingestion points: User prompts regarding logic states and UI design patterns are processed within the branches defined in SKILL.md and detailed in LOGIC.md and UI.md.
Boundary markers: The skill does not provide instructions for the agent to use delimiters or specific safety markers when interpolating user requirements into the generated prototype code.
Capability inventory: The skill allows the agent to generate and write executable files (HTML, JavaScript, React components) and suggests the execution of shell commands (pnpm, python, bun) as described in Rule 2 of SKILL.md.
Sanitization: No sanitization, escaping, or validation requirements are specified for the user-provided input before it is used to generate executable code.
[COMMAND_EXECUTION]: SKILL.md instructs the agent to define and suggest command-line operations for the user to run the generated prototypes (e.g., pnpm <name>, python <path>). While intended for development convenience, this pattern could lead to command injection if the agent derives filenames or paths from unvalidated user input.
[DYNAMIC_EXECUTION]: The skill's primary function is the dynamic generation of executable code, including logic modules and UI components, based on runtime prompts. While this is the intended purpose, generating code from untrusted high-level descriptions is a form of dynamic execution that inherits risks associated with the input's source.

prototype