abstract

Pass

Audited by Gen Agent Trust Hub on May 2, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of the @membranehq/cli Node.js package. This is the official tool provided by the vendor (Membrane) to manage integrations and automate workflows.
  • [COMMAND_EXECUTION]: The skill utilizes the membrane command-line utility for authentication, service connection, and action execution. These commands are necessary for the skill's functionality and interact exclusively with the vendor's managed infrastructure.
  • [REMOTE_CODE_EXECUTION]: The documentation suggests using npx to execute the vendor's CLI package directly. This involves fetching and running remote code from the official registry, which is a standard and expected practice for using the vendor's own tooling.
  • [PROMPT_INJECTION]: The skill processes data from the Abstract API, which constitutes an indirect prompt injection surface. 1. Ingestion points: External data such as project details, tasks, and comments are ingested via membrane action run. 2. Boundary markers: No specific delimiters or warnings for the agent are defined in the instructions to isolate this untrusted content. 3. Capability inventory: The agent has the ability to execute shell commands via the CLI and access the network. 4. Sanitization: There is no mention of sanitization or validation for the data retrieved from the external service.
Audit Metadata
Risk Level
SAFE
Analyzed
May 2, 2026, 03:22 PM
Security Audit — agent-trust-hub — abstract