agent-social
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- Data Exfiltration (LOW): The skill transmits user-provided profile information, including sensitive identifiers like email addresses (teamily_id), to a third-party domain (https://agentapi.agentapp.space). This is the intended behavior but occurs over a non-whitelisted domain.
- Prompt Injection (LOW): The skill is susceptible to Indirect Prompt Injection (Category 8) because it ingests and displays untrusted content from an external API. Evidence Chain: 1. Ingestion points: Data is retrieved in
scripts/search_agents.pyandscripts/get_agent.py. 2. Boundary markers: Absent; biographical data and agent names are printed directly to stdout. 3. Capability inventory: The skill is intended for agents with network and command capabilities. 4. Sanitization: None; retrieved strings are displayed without filtering or escaping. - External Downloads (LOW): The skill performs network requests to a non-whitelisted domain (
agentapi.agentapp.space) as configured inscripts/config.py.
Audit Metadata