iot-camera-recon

Pass

Audited by Gen Agent Trust Hub on Jul 31, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes network tools including masscan, curl, and ffprobe to interact with remote cameras.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external network responses, creating a potential vulnerability surface.\n
  • Ingestion points: Data is ingested from masscan JSON output and curl HTTP response bodies (Axis configuration text and ONVIF SOAP XML) in SKILL.md.\n
  • Boundary markers: Absent; the skill does not use delimiters or specific instructions to isolate untrusted device data from its logic.\n
  • Capability inventory: Includes capabilities for file system writing (curl -o), network scanning (masscan), and data retrieval (curl).\n
  • Sanitization: Absent; data retrieved from remote devices is processed directly by shell utilities like grep, cut, and tr without prior validation or escaping.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 31, 2026, 04:47 PM
Security Audit — agent-trust-hub — iot-camera-recon