redis-state-management
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [SAFE]: The skill implements established industry standards for high-performance state management. It provides detailed guidance on connection pooling, high availability via Redis Sentinel, and the use of SSL/TLS for secure communication.
- [INDIRECT_PROMPT_INJECTION]: The skill provides patterns for processing and storing external data, such as chat messages, user activity logs, and job queue payloads. This creates a surface for indirect prompt injection where untrusted data could be retrieved and interpreted by the agent.
- Ingestion points: Chat message history in EXAMPLES.md, job queue payloads in EXAMPLES.md, and user activity tracking in SKILL.md.
- Boundary markers: None identified in the provided templates or instructions.
- Capability inventory: Network connectivity to Redis servers, ability to write and read persistent state.
- Sanitization: The skill uses JSON serialization for data structures but does not include sanitization logic to detect or neutralize natural language instructions embedded within the data values.
- [METADATA_POISONING]: The skill contains a metadata inconsistency; the author is listed as 'manutej', but the included VALIDATION.md file attributes the skill's creation to 'Claude Code (Anthropic)'. This conflicting attribution is a minor documentation issue.
Audit Metadata