delighted

Pass

Audited by Gen Agent Trust Hub on Apr 28, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Downloads the official Membrane CLI tool (@membranehq/cli) from the public NPM registry. This tool is necessary for interacting with the Membrane platform and is a trusted resource from the vendor.
  • [COMMAND_EXECUTION]: Uses the 'membrane' CLI to perform authenticated operations, including creating connections, searching for available actions, and running data workflows. These operations are restricted to the authenticated user's environment.
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it processes external data from Delighted.
  • Ingestion points: Delighted survey responses and person data ingested via CLI commands.
  • Boundary markers: None explicitly defined in the skill instructions.
  • Capability inventory: File system access through CLI installation, network access for API calls, and action execution capabilities.
  • Sanitization: Relies on the underlying LLM's safety filters and the Membrane platform's execution environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 28, 2026, 04:46 PM
Security Audit — agent-trust-hub — delighted