The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses highly sensitive local files stored in ~/.proficiently/, including linkedin-contacts.csv and the user's resume and preferences. While these are used for the stated purpose of job matching, the automated processing of personal network data and professional history from the local filesystem represents a data exposure surface.
[COMMAND_EXECUTION]: The skill utilizes powerful MCP tools for browser automation (tabs_create_mcp, navigate, get_page_text) and the Task tool to execute parallel subagents. These capabilities allow the agent to interact with arbitrary external websites and spawn additional execution contexts.
[PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection. In Step 3, the agent navigates to external careers pages and job boards, extracting content via get_page_text. Maliciously crafted job descriptions on these third-party sites could contain instructions designed to influence the agent's behavior, such as manipulating the job match results or attempting to exfiltrate cached data. There are no explicit sanitization routines or boundary markers defined in the instructions to protect against such embedded commands.

network-scan