handover-test-continuity
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [DATA_EXFILTRATION]: The skill includes strong negative constraints, explicitly ordering the agent to never request, display, log, or publish credentials, and to avoid including personal data in created artifacts.
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates data movement between distinct authenticated agents, creating a potential ingestion surface. 1. Ingestion points:
handover.read_artifactandhandover.getinSKILL.md. 2. Boundary markers: Session markers (e2e-mcp-...) are used for tracking, though specific text delimiters are not defined. 3. Capability inventory:handover.create,handover.continue, andhandover.annotateare utilized to manage state. 4. Sanitization: It requires identity verification viahandover.whoamiand artifact verification against the expected session state. - [EXTERNAL_DOWNLOADS]: References documentation and workflow examples from the official handover.sh domain to provide schemas and technical guidance.
Audit Metadata