sensorkit
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of highly sensitive and untrusted data streams from user interactions (keyboard metrics, speech transcriptions, messages usage) without providing security guidance.
- Ingestion points: Untrusted data enters the agent's context through
SRKeyboardMetrics(SKILL.md, references/usage-and-environment-samples.md),SRMessagesUsageReport(references/usage-and-environment-samples.md), andSRSpeechMetrics(references/speech-face-and-health-samples.md). - Boundary markers: The documentation and code snippets do not implement or recommend the use of boundary markers or delimiters when processing these raw sensor streams.
- Capability inventory: The skill provides the mechanism to fetch biometric, interaction, and environmental data via the
SRSensorReader.fetchmethod across multiple sensors. - Sanitization: There is no mention of sanitizing, escaping, or filtering the external content derived from user typing, speech recognition results, or messaging events before the agent processes the information.
Audit Metadata