adk-observability-safety

Wire ADK 2.0 logging, tool monitoring, safety policies, and auth

Use this skill to configure ADK 2.0 authentication — gating who can call the agent and which tools require which scopes. Triggers on: "ADK auth", "secure ADK agent", "ADK authentication", "ADK OAuth", "ADK API key auth", "ADK service account", "tool-level auth ADK", "ADK bearer token", "ADK identity-aware proxy". Generates auth scheme definitions, per-tool scope requirements, and integration with IAP / OAuth / API key headers.

Use this skill to wire structured logging and callbacks into an ADK 2.0 agent — capture every model call, tool call, state change for debugging and audit. Triggers on: "ADK logging", "ADK callbacks", "log ADK events", "ADK observability", "trace ADK agent", "structured logs ADK", "ADK debug logging". Generates callback handlers wired to Cloud Logging / OpenTelemetry / structured stdout for production observability.

Use this skill to add safety policies to an ADK 2.0 agent — input filtering, output redaction, prompt-injection detection, blocked-topic enforcement. Triggers on: "ADK safety", "ADK guardrails", "block harmful content ADK", "ADK content filter", "prompt injection ADK", "PII redaction ADK", "ADK policy enforcement", "safe agent output". Generates pre/post callbacks that screen inputs and outputs against policy rules and either redact, refuse, or escalate.

Use this skill to set up a tool-call monitoring dashboard for ADK 2.0 agents — track tool latency, error rate, usage frequency, and per-tool cost. Triggers on: "ADK tool monitoring", "ADK dashboard", "monitor ADK tool calls", "tool latency ADK", "ADK metrics", "Grafana ADK", "Cloud Monitoring ADK", "ADK tool stats". Generates metric exporters and a reference dashboard config.