The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes unstructured user artifacts (proposals, design docs, notes) which constitutes an ingestion point for untrusted data.
Ingestion points: Processes files in the {document_directory} provided by the user.
Boundary markers: The skill does not explicitly define markers but uses a structured 'Phase' approach to separate data analysis from decision making.
Capability inventory: The skill uses write_to_file to save a strategy document and ask_question for user interaction. It does not execute shell commands or perform network operations.
Sanitization: No explicit sanitization of input text is mentioned, though it is processed through a synthesis model rather than direct execution.

distilling-strategies-interactively