infisical-gateway

Warn

Audited by Gen Agent Trust Hub on Aug 26, 2026

Risk Level: MEDIUMPRIVILEGE_ESCALATIONPERSISTENCECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [PRIVILEGE_ESCALATION]: The skill provides instructions for installing the gateway as a system service using sudo, which grants the application root-level privileges on the host system to perform system-level operations.
  • Evidence: sudo infisical gateway systemd install command in references/deployment-and-networking.md.
  • [PERSISTENCE]: The skill utilizes systemd to ensure the gateway service automatically starts and remains active across system reboots, creating a persistent presence on the host.
  • Evidence: Deployment steps in references/deployment-and-networking.md that configure the service via the systemd install command.
  • [COMMAND_EXECUTION]: The skill contains multiple shell commands for the agent or user to execute for enrollment, starting, and managing the gateway lifecycle.
  • Evidence: Commands such as infisical gateway start and infisical gateway systemd install in references/deployment-and-networking.md.
  • [INDIRECT_PROMPT_INJECTION]: The gateway architecture creates a communication bridge where an external platform (Infisical) can send requests into private internal networks, introducing a surface for potentially malicious instructions to reach sensitive internal services.
  • Ingestion points: The gateway receives proxied requests from the Infisical platform through an outbound SSH reverse tunnel via a relay server (SKILL.md).
  • Boundary markers: The system uses certificate-based SSH authentication and double encryption to protect the traffic channel (SKILL.md).
  • Capability inventory: The gateway has the capability to interact with private databases, directories (LDAP), internal APIs, and Hardware Security Modules (HSMs) using TCP or HTTP protocols (references/pools-and-consumers.md).
  • Sanitization: Traffic is double-encrypted between the platform and relay; however, the gateway ultimately translates these platform-originated requests into operations on the local internal network (references/deployment-and-networking.md).
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 26, 2026, 11:05 PM
Security Audit — agent-trust-hub — infisical-gateway