docker-hub

Warn

Audited by Snyk on Apr 23, 2026

Risk Level: MEDIUM
Full Analysis

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

  • Third-party content exposure detected (high risk: 0.90). The skill explicitly instructs the agent to fetch and proxy content from Docker Hub (e.g., via "membrane action run" and "membrane request CONNECTION_ID /path/to/endpoint"), which exposes user-generated/untrusted repository data (READMEs, tags, descriptions) that the agent is expected to read and that could materially influence subsequent tool use or decisions.

Issues (1)

W011
MEDIUM

Third-party content exposure detected (indirect prompt injection risk).

Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 23, 2026, 05:58 AM
Issues
1