dropbox

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of the @membranehq/cli package globally via npm to interact with the Membrane platform and manage Dropbox connections.- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources by listing folder contents, searching files, and retrieving metadata from Dropbox.
  • Ingestion points: Data enters the context through actions like list-folder-contents, get-metadata, and search-files in SKILL.md.
  • Boundary markers: None explicitly defined in the skill instructions to delimit file content from system prompts.
  • Capability inventory: The skill allows network requests and action execution via the membrane CLI.
  • Sanitization: No explicit sanitization or filtering of Dropbox content is described in the provided file.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 03:16 PM
Security Audit — agent-trust-hub — dropbox