Skills
Navigate
skills/kevinbadi/ai-os-skills/linkedin-connection-agent/Gen Agent Trust Hub

linkedin-connection-agent

Fail

Audited by Gen Agent Trust Hub on Apr 8, 2026

Risk Level: HIGHCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: A hardcoded API key for PhantomBuster was detected in the reference documentation.
  • Evidence: PHANTOMBUSTER_API_KEY=QGUtgLhzz2i437bWBM6UN0MBoagJoFR3OX7UHT7dKSI in references/linkedin_connection_agent.md.
  • [EXTERNAL_DOWNLOADS]: Fetches automated activity results and structured LinkedIn profile data from PhantomBuster's official API and AWS S3 storage endpoints.
  • [DATA_EXFILTRATION]: Retrieves LinkedIn profile details (names, profile URLs, and company information) from a remote source and stores them in an external PostgreSQL database.
  • [PROMPT_INJECTION]: The skill ingests untrusted LinkedIn profile data from PhantomBuster, creating a surface for indirect prompt injection.
  • Ingestion points: profiles = json_response.json() in scripts/linkedin_connection_agent.py and scripts/modal_linkedin_connections.py.
  • Boundary markers: Absent.
  • Capability inventory: Database write access via psycopg2 and network request capabilities via requests.
  • Sanitization: Absent; no validation or escaping of profile content was found before processing.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Apr 8, 2026, 08:20 AM
Security Audit — agent-trust-hub — linkedin-connection-agent