Skills
Navigate
skills/gemini-cli-extensions/data-agent-kit-starter-pack/google-cloud-storage-basics/Gen Agent Trust Hub

google-cloud-storage-basics

Pass

Audited by Gen Agent Trust Hub on Sep 6, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill allows the agent to read and process object content from Cloud Storage buckets (e.g., via read_object, download_object, and gcsfuse mounts). If these buckets contain untrusted data, it could lead to indirect prompt injection attacks.
  • Ingestion points: Objects are read via the read_object tool in references/mcp-usage.md and via file system operations on gcsfuse mounts described in references/gcsfuse.md.
  • Boundary markers: The instructions do not specify the use of delimiters or "ignore embedded instructions" prompts when processing object content.
  • Capability inventory: The skill utilizes the gcloud CLI for bucket and object management, and provides tools for writing objects, copying/moving files, and managing IAM policies as detailed in SKILL.md and references/mcp-usage.md.
  • Sanitization: No explicit sanitization or validation of the object content is mentioned before the agent processes it.
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions for downloading official Google tools and configuration.
  • Evidence: Downloads gcsfuse from packages.cloud.google.com (File: references/gcsfuse.md) and the MCP Toolbox binary from storage.googleapis.com (File: references/mcp-usage.md). These are official, well-known service domains.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 6, 2026, 01:34 PM
Security Audit — agent-trust-hub — google-cloud-storage-basics