upstash-redis-start

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates reading data from an external Redis instance, which could be populated with untrusted content designed to influence agent behavior.
  • Ingestion points: Values retrieved using commands like GET, LRANGE, and ZREVRANGE in SKILL.md.
  • Boundary markers: Absent; there are no specific instructions for the agent to treat database content as untrusted or to use delimiters.
  • Capability inventory: The skill uses curl to perform network requests to the Upstash REST API (SKILL.md).
  • Sanitization: Absent; the skill body does not specify any validation, filtering, or escaping logic for the data retrieved from Redis.
  • [COMMAND_EXECUTION]: The skill instructs the agent to use curl to interact with Upstash's provisioning and database endpoints.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 05:42 AM
Security Audit — agent-trust-hub — upstash-redis-start