Skills
skills/nikiandr/goose-skills/job-posting-intent/Gen Agent Trust Hub

job-posting-intent

Fail

Audited by Gen Agent Trust Hub on Mar 28, 2026

Risk Level: HIGHCREDENTIALS_UNSAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: Hardcoded authentication tokens (JWTs) for the 'rube.app' service are present in multiple files, including scripts/create_sheet.py, scripts/create_sheet_mcp.py, and scripts/search_jobs.py. These tokens grant access to the Rube MCP and Composio integrations.
  • [REMOTE_CODE_EXECUTION]: The skill generates Python code strings at runtime and sends them to a remote server (rube.app/mcp) for execution via the RUBE_REMOTE_WORKBENCH tool. This pattern of remote code generation and execution is used in scripts/search_jobs.py and scripts/create_sheet_mcp.py to manage Google Sheets operations.
  • [COMMAND_EXECUTION]: The skill uses the RUBE_REMOTE_WORKBENCH and Composio execution endpoints to run arbitrary logic and tools on a remote platform.
  • [REMOTE_CODE_EXECUTION]: Data retrieved from external LinkedIn job postings via Apify is interpolated into the Python scripts executed on the remote workbench. While json.dumps is used for formatting, the overall architecture of executing code constructed from untrusted external data represents a potential injection vulnerability where malicious job posting content could influence the remote execution logic.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Mar 28, 2026, 11:42 AM