retain
Pass
Audited by Gen Agent Trust Hub on Apr 29, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, data exfiltration attempts, or unauthorized command executions were found. The skill follows best practices for retention strategy and incorporates explicit warnings against dark patterns.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external feedback data (from the 'Voice' agent), which represents a theoretical attack surface for indirect prompt injection. However, the instructions focus on qualitative analysis and interpretation rather than execution of data content, and the risk is considered negligible within the provided operational context.
- Ingestion points: Feedback data and churn reasons from the 'Voice' agent (SKILL.md).
- Boundary markers: Not explicitly defined for external data, but the skill scope is limited to strategy and reporting.
- Capability inventory: No subprocess calls or network operations; capabilities are limited to analytical report generation.
- Sanitization: Not explicitly mentioned, but recommendations are restricted to behavioral evidence.
Audit Metadata