The Agent Skills Directory

[SAFE]: The skill instructions define a legitimate development workflow for creating isolated logic and UI prototypes. No unauthorized network activity, data exfiltration, or malicious persistence mechanisms were detected.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-defined questions to generate executable code, which presents a surface for potential injection of malicious logic within the generated artifacts.
Ingestion points: User-provided questions and logic descriptions used to generate LOGIC.md and UI.md components.
Boundary markers: The skill uses specific directory isolation (.context/prototypes/) but does not explicitly define prompt delimiters for the user input.
Capability inventory: The skill allows writing files to the local filesystem and executing generated code via a /dogfood command.
Sanitization: No explicit sanitization or validation of the user-provided logic is described; the skill relies on the agent's internal reasoning during code generation.
[DYNAMIC_EXECUTION]: The instructions in SKILL.md call for a /dogfood step to verify the generated prototype. This involves executing newly created scripts or HTML files to check logic or UI behavior, which is a standard part of the development process intended by the skill.

prototype