The Agent Skills Directory

[SAFE]: The skill implements a secure human-in-the-loop pattern. It mandates that the agent summarize all fields and obtain explicit user approval before executing any write-capable actions (e.g., create_draft, update, confirm_preview).- [INDIRECT_PROMPT_INJECTION]: The skill handles external data, such as existing job postings and market rate insights. Potential risks are mitigated by instructions to never reuse old terms blindly, to use get_tool_help for valid formats, and to require confirmation for how skills and qualifications are interpreted before saving or publishing. The skill also explicitly references handling untrusted participant text according to server rules.

write-job-post