northflank

Pass

Audited by Gen Agent Trust Hub on May 1, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the user to install the @membranehq/cli package globally via npm. This is a documented vendor resource associated with the Membrane platform.
  • [COMMAND_EXECUTION]: All interaction with the Northflank API is mediated through the membrane CLI tool. This includes authentication, action discovery, and running operations. The skill explicitly advises against manual credential handling, favoring the platform's internal security model.
  • [PROMPT_INJECTION]: The skill demonstrates an indirect prompt injection surface where natural language strings are passed to membrane action list --intent and membrane action create. While these fields ingest untrusted input that could influence which actions are found or generated, the risk is mitigated by the platform's state management and the agent's expected review of action schemas before execution.
  • Ingestion points: natural language queries in membrane action list and descriptions in membrane action create (SKILL.md).
  • Boundary markers: None explicitly defined in command templates.
  • Capability inventory: Execution of Northflank management operations via shell commands (SKILL.md).
  • Sanitization: Not explicitly documented; relies on the Membrane platform's internal validation during action generation.
Audit Metadata
Risk Level
SAFE
Analyzed
May 1, 2026, 09:23 PM
Security Audit — agent-trust-hub — northflank