The Agent Skills Directory

[SAFE]: No security threats were identified in this skill. The skill follows standard operational patterns for its stated purpose.
[DATA_ACCESS]: The skill requests read/write permissions for specific directories: ~/.claude/skills/** (its own code) and ~/.proficiently/** (its local data store for resumes and job postings). This is consistent with its functionality as a job application tool.
[EXTERNAL_TOOLS]: It utilizes 'Claude in Chrome' MCP tools to fetch job descriptions from URLs provided by the user. This is a legitimate use of browser capabilities for data gathering.
[INDIRECT_PROMPT_INJECTION]: While the skill processes external job postings (which are untrusted data), it does so to extract requirements. The output is a cover letter presented to the user for review before any further action is taken, which serves as a human-in-the-loop safety measure.

cover-letter