adk-observability-guide
Pass
Audited by Gen Agent Trust Hub on Mar 7, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches documentation from Google's official documentation site (google.github.io/adk-docs/).\n- [COMMAND_EXECUTION]: Provides standard commands for installing the agentops package and using the agent-starter-pack tool for project scaffolding.\n- [DATA_EXFILTRATION]: Documents the configuration of telemetry and logging to Google Cloud Storage and BigQuery, including environment variables for bucket names and privacy settings.\n- [PROMPT_INJECTION]: Identifies a surface for indirect prompt injection via the collection of GenAI interaction logs.\n
- Ingestion points: Untrusted user prompts and model responses are captured as described in the Prompt-Response Logging section of SKILL.md.\n
- Boundary markers: The skill highlights the use of the NO_CONTENT privacy mode as a default behavior.\n
- Capability inventory: The guide mentions tools like bq query, gsutil, and gcloud for verification in references/cloud-trace-and-logging.md.\n
- Sanitization: Recommends the NO_CONTENT setting to log only metadata (tokens, model name, timing) instead of full prompt/response content in production.
Audit Metadata