vext

Pass

Audited by Gen Agent Trust Hub on Apr 29, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads and installs the @membranehq/cli package from the NPM registry. This package is the official command-line tool for the Membrane platform.
  • [COMMAND_EXECUTION]: Uses the membrane CLI to perform operations such as login, connection management, and action execution. These commands are run in the local shell environment.
  • [REMOTE_CODE_EXECUTION]: Provides functionality to dynamically generate new actions on the Membrane platform via membrane action create. This results in the server-side creation and subsequent execution of code.
  • [DATA_EXPOSURE_AND_EXFILTRATION]: Demonstrates good security posture by explicitly advising against requesting or storing raw API keys, instead utilizing the platform's connection management to handle the authentication lifecycle securely.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data via the membrane action run command, which presents an attack surface for indirect prompt injection.
  • Ingestion points: Data enters the system via the --input parameter of the membrane action run command in SKILL.md.
  • Boundary markers: The instructions do not define boundary markers or delimiters for untrusted data.
  • Capability inventory: The skill can execute various platform actions and shell commands defined in SKILL.md.
  • Sanitization: No input validation or sanitization mechanisms are implemented for the data passed to the CLI tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 29, 2026, 06:50 PM
Security Audit — agent-trust-hub — vext