Skills
skills/gooseworks-ai/gooseworks-skills/job-posting-intent/Gen Agent Trust Hub

job-posting-intent

Fail

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: HIGHCREDENTIALS_UNSAFEREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: Hardcoded authorization token (RUBE_TOKEN) found in scripts/create_sheet.py, scripts/create_sheet_mcp.py, and scripts/search_jobs.py. This JSON Web Token (JWT) provides access to the Rube/Composio platform and is shared across all users of the skill.
  • [REMOTE_CODE_EXECUTION]: The skill employs the RUBE_REMOTE_WORKBENCH tool to execute Python code constructed at runtime on a remote server (rube.app). This code includes data retrieved from job postings.
  • [EXTERNAL_DOWNLOADS]: The skill communicates with external APIs including the Apify proxy (app.gooseworks.ai) and the Rube platform (rube.app).
  • [PROMPT_INJECTION]: The skill has an indirect prompt injection surface.
  • Ingestion points: LinkedIn job descriptions are fetched via the Apify actor in scripts/search_jobs.py.
  • Boundary markers: None. External job data is concatenated and processed without delimiters or instructions for the agent to ignore embedded instructions.
  • Capability inventory: The skill can perform network requests and execute code on a remote workbench.
  • Sanitization: Scraped data is serialized via json.dumps before being embedded in the remote workbench script, offering basic structural protection but no defense against malicious content.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Jul 10, 2026, 03:06 PM
Security Audit — agent-trust-hub — job-posting-intent