The Agent Skills Directory

[SAFE]: The skill instructions strictly define a workflow for document generation and do not contain any patterns suggestive of bypassing security protocols or overriding safety filters.
[DATA_EXFILTRATION]: All file operations are localized to the 'tasks/' directory using the Write tool. There are no network-capable tools or instructions that would allow for the transmission of data to external servers.
[PROMPT_INJECTION]: The skill incorporates user input into a Markdown document as part of its primary function. While this creates a standard surface for indirect prompt injection, the skill's lack of high-risk capabilities (such as shell execution or network access) ensures that any injected content remains within a non-executable document context.

prd