diffbot

Pass

Audited by Gen Agent Trust Hub on Apr 28, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches the Membrane CLI tool (@membranehq/cli) from the official NPM registry to facilitate interaction with the Membrane platform.
  • [COMMAND_EXECUTION]: Utilizes the membrane command-line interface to perform authentication, manage connections, and execute data extraction tasks, including the dynamic creation of actions.
  • [PROMPT_INJECTION]: The skill ingests untrusted data from the web through Diffbot's extraction capabilities (articles, products, discussions, etc.) which are returned via CLI output.
  • Ingestion points: External web content processed by Diffbot and returned via 'membrane action run' in SKILL.md.
  • Boundary markers: None provided; the instructions do not specify the use of delimiters or warnings to ignore instructions within the extracted content.
  • Capability inventory: The skill can list, create, and execute actions via the 'membrane' CLI in SKILL.md.
  • Sanitization: None detected; the skill does not explicitly sanitize or validate the extracted data before it is processed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 28, 2026, 10:02 AM
Security Audit — agent-trust-hub — diffbot