cloud-trace-querying
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- Command Execution: The skill utilizes shell commands via
gcloudandsubprocesscalls to interact with Google Cloud services and manage local execution environments. These are standard operations for the skill's intended purpose of cloud observability. - Dynamic Environment Management: The skill includes a sophisticated Python virtual environment manager that handles dependency installation and process execution. It uses file-based locking and automated cleanup to ensure a stable and isolated runtime.
- Indirect Prompt Injection Surface: The skill ingests untrusted telemetry data from Cloud Trace and Cloud Logging. To mitigate risks, it uses a sanitization pattern where outputs are wrapped in randomized
<untrusted_telemetry>tags. Ingestion occurs in the main tool scripts (e.g.,fetch_related_logs), and the skill's capabilities include GCP API interactions and local file writing. The presence of these boundary markers significantly reduces the attack surface. - Secure Credential Handling: The skill relies on standard Google Cloud authentication mechanisms and Application Default Credentials, avoiding the use of hardcoded secrets or unsafe credential storage.
Audit Metadata