fingerprinting-server-software

Pass

Audited by Gen Agent Trust Hub on Jul 24, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external web servers which could be crafted to influence the agent's behavior during the analysis phase.
  • Ingestion points: External HTTP response headers (resp.headers) and response bodies (resp.text) are parsed by scripts/fingerprint_server.py.
  • Boundary markers: None identified. The script extracts data from these sources and includes them directly in finding reports without delimiters meant to isolate untrusted content from the agent's instructional context.
  • Capability inventory: The skill can perform network requests (lib.http_client) and write reports to the filesystem if the --output flag is used.
  • Sanitization: Header values are truncated to 80 characters in the finding titles, but the full raw value is included in the evidence and details, providing a vector for potentially malicious content to enter the agent's context.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 24, 2026, 06:34 AM
Security Audit — agent-trust-hub — fingerprinting-server-software