The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input (feature descriptions) to generate a document. This represents a potential indirect injection surface if the resulting document is later processed by other automated tools. However, the skill incorporates a 'HARD-GATE' that strictly forbids the inclusion of implementation code, SQL, or technical signatures, requiring the output to remain in natural language, which significantly mitigates risks to downstream agents.
[SAFE]: The skill operates entirely within the local environment, writing structured Markdown files to a designated directory. No network operations, hardcoded credentials, or dynamic code execution patterns were found.

create-prd