The Agent Skills Directory

[SAFE]: The skill is composed entirely of natural language instructions for the AI agent and contains no scripts, binaries, or automated subprocess commands.
[PROMPT_INJECTION]: The skill processes untrusted user input (the 'messy ask') to generate a prompt deliverable. While it lacks automated sanitization, the iterative interview workflow acts as a natural boundary where the agent confirms intent before final output, minimizing the risk of accidental malicious instruction execution.
Ingestion points: The argument-hint field and subsequent user responses during the clarification workflow.
Boundary markers: No explicit prompt delimiters are defined for the user input.
Capability inventory: The skill suggests the agent can execute the final prompt or save it to a file, but these actions are triggered by user choice at the end of the workflow rather than the skill itself.
Sanitization: The instructions do not define specific escaping or validation rules for user-supplied strings.
[DATA_EXPOSURE]: The instructions direct the agent to analyze the local codebase and configuration files (such as .env, README.md, or project structure) to gather context. This is intended for environment awareness to improve prompt quality and does not include logic for external data exfiltration.

clarify