The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-defined design questions to generate functional code, creating an attack surface where adversarial input could attempt to influence the logic of the generated scripts.
Ingestion points: User requirements for prototypes are ingested to determine the shape of the code.
Boundary markers: Instructions restrict prototypes to answering a single question and emphasize that the code is throwaway.
Capability inventory: The agent is tasked with writing code files and identifying shell commands to run them.
Sanitization: No specific input sanitization is defined within the skill instructions.
[DYNAMIC_EXECUTION]: The skill instructs the agent to generate new source code, such as terminal-based state machines or React/web components, which are intended to be executed immediately.
Evidence: references/logic.md and references/ui.md provide templates and rules for creating interactive code modules at runtime.
[COMMAND_EXECUTION]: The skill directs the agent to determine and suggest shell commands to run the generated prototypes using the project's existing infrastructure.
Evidence: SKILL.md rule 2 specifies "One command to run, via the project's existing task runner — the user starts it without thinking."

prototype