gcp-playbooks

Pass

Audited by Gen Agent Trust Hub on Aug 25, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides a deterministic set of gcloud and kubectl commands for infrastructure discovery, project management, and incident mitigation (e.g., rollbacks and scaling), which align with the skill's stated SRE purpose.- [EXTERNAL_DOWNLOADS]: The included script report_errors_by_revision.py specifies dependencies on established libraries such as google-api-python-client, google-auth, and python-dateutil from official registries.- [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection due to its ingestion of error log data.
  • Ingestion points: The report_errors_by_revision.py script retrieves error messages and stack traces from the GCP Error Reporting API.
  • Boundary markers: The script does not utilize specific delimiters or isolation markers when outputting the retrieved error messages to the agent context.
  • Capability inventory: The skill has the capability to execute administrative shell commands and modify project configurations.
  • Sanitization: Error messages are displayed without sanitization or validation of their content.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 25, 2026, 01:11 AM
Security Audit — agent-trust-hub — gcp-playbooks