gcloud

Pass

Audited by Gen Agent Trust Hub on Jun 28, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides structured workflows for executing gcloud CLI commands. It implements several defensive measures, including mandatory command validation using gcloud help, a prohibition on shell operators (pipes, redirection), and a denylist of destructive operations (e.g., IAM modifications, resource deletion) to prevent autonomous damage.
  • [EXTERNAL_DOWNLOADS]: The skill references official Google Cloud documentation and installation guides. These links point to cloud.google.com, which is a well-known and trusted service.
  • [PROMPT_INJECTION]: Potential surface for indirect prompt injection via tool output ingestion. 1. Ingestion points: Command output from gcloud help and gcloud list (SKILL.md). 2. Boundary markers: Absent. 3. Capability inventory: Execution of gcloud CLI commands. 4. Sanitization: Absent. The risk is mitigated by the skill's mandate to use structured formatting to reduce the data volume ingested.
  • [CREDENTIALS_UNSAFE]: The skill discusses authentication mechanisms including service account keys and impersonation. It correctly identifies the security risks of local JSON key files and recommends service account impersonation as a safer alternative for local development.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 28, 2026, 05:51 PM
Security Audit — agent-trust-hub — gcloud