jobgpt

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (low risk: 0.10). Exclusion: The required runtime workflow described calls JobGPT MCP tools based on user-provided job search filters/URLs (e.g., import_job_by_url) but the skill documentation does not specify ingesting any outsider-authored free text into the LLM without first selecting an item (it’s centered on importing/processing specific job pages rather than monitoring a feed/queue or reading arbitrary text sources at runtime).

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly requires and configures an external MCP server at runtime using the URL https://mcp.6figr.com/mcp, which the agent will call to execute tool actions and thus can directly control prompts/behavior and run remote code.