google-cloud-storage-basics

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s required runtime workflow can read outsider-authored free text via the Cloud Storage MCP remote server tool read_object (and similar “read text/PDF/image” operations) at runtime when an agent issues MCP calls like tools/call/read_object, causing the model to ingest object contents supplied by the bucket’s submitter.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). I flagged runtime downloads that fetch and install/execute remote binaries or keys (curl -L -o toolbox https://storage.googleapis.com/mcp-toolbox-for-databases/VERSION/linux/amd64/toolbox and curl https://packages.cloud.google.com/apt/doc/apt-key.gpg) because they retrieve remote executable content or package keys during setup which will be executed/used locally.