skills/terminalskills/skills/rabbitmq/Gen Agent Trust Hub

rabbitmq

Warn

Audited by Gen Agent Trust Hub on Aug 31, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The SKILL.md file contains hardcoded default credentials ('admin' and 'secret') within both the Docker Compose configuration and the Node.js connection string. While these are documented as local development defaults for localhost, they represent hardcoded secrets in code templates.
  • [INDIRECT_PROMPT_INJECTION]: The skill demonstrates consuming and processing messages from a RabbitMQ queue, which serves as a surface for indirect prompt injection.
  • Ingestion points: The startWorker function in SKILL.md (task-consumer.ts) uses ch.consume to ingest raw message content from an external queue into the application context.
  • Boundary markers: Absent; the payload is parsed directly as JSON, and the logic lacks delimiters or system instructions to ignore embedded commands within the message body.
  • Capability inventory: The consumer script handles tasks that can trigger external actions, such as sendEmail, or other business logic that may be manipulated by injected instructions.
  • Sanitization: The example does not include validation or sanitization logic for the message payload before it is passed to the handler function.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 31, 2026, 01:20 AM
Security Audit — agent-trust-hub — rabbitmq