The Agent Skills Directory

[PROMPT_INJECTION]: An indirect prompt injection surface was identified within the documentation workflow.
Ingestion points: User-provided business ideas, feature requirements, and persona details collected during Step 1 are directly incorporated into output files.
Boundary markers: The skill lacks explicit delimiters or instructions to the agent to treat user-provided input strictly as data, increasing the risk of the agent following instructions embedded within the requirements.
Capability inventory: The skill is granted Read, Write, Edit, Grep, and Glob tools, enabling it to persist potentially malicious content to the /docs/prd/ directory.
Sanitization: No validation or escaping of user input is performed before writing data to the markdown documents.

prd-documentation