external-source-adapter
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill instructions promote security best practices, such as avoiding hard-coded credentials and explicitly requiring environment variables for sensitive configuration like MQTT authentication and root certificates.
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the mapping of external telemetry data (MAVLink, ROS 2, NMEA, CAN, UWB/RTLS) which represents a potential ingestion surface for untrusted data.
- Ingestion points: External telemetry streams from physical or simulated devices.
- Boundary markers: Data is constrained by a canonical JSON contract and validated against specific schemas (
metadata.schema.json). - Capability inventory: The skill involves writing Python scripts that utilize MQTT publishing via the platform's
scene_common.mqtt.PubSubutility. - Sanitization: Payloads are structured according to fixed JSON schemas, minimizing the risk of arbitrary instruction execution through the telemetry pipeline.
Audit Metadata