Kipt

Your prompts and skills, one keystroke away.

Explore

google-agents-cli-observability skill: trace and log deployed agents

The agents-cli observability skill from google/agents-cli guides coding agents through Cloud Trace, prompt-response logging, BigQuery analytics and integrations.

By the Kipt team · Updated 26 October 2026

Source: google/agents-cli · Maintainer: Google · License: Apache-2.0 · Checked: 26 October 2026

What is the google-agents-cli-observability skill?

google-agents-cli-observability is an Agent Skill that teaches a coding agent how to monitor AI agents built and deployed with the agents-cli command. It covers Cloud Trace, prompt-response logging, the BigQuery Agent Analytics plugin, third-party platforms and common problems.

It is one of seven skills in google/agents-cli, Google's "CLI and skills for building agents on Gemini platform." In the workflow skill it is Phase 7, after deploy. The SKILL.md description says not to use it for deployment setup or API code patterns.

As of 26 October 2026, its skills.sh page listed about 511,700 installs (snapshot, 26 October 2026). The SKILL.md metadata lists Google as author and version 1.11.0.

What does it do?

The SKILL.md is organized around four tiers and tells the agent to ask you which ones you need. Cloud Trace is always on and the rest are additive.

Tier What it captures Scope Default
Cloud Trace Execution flow, latency and errors as OpenTelemetry spans All templates, all environments Always on
Prompt-response logging GenAI interactions to GCS, BigQuery and Cloud Logging Scaffolded ADK Python projects Off locally, on when deployed
BigQuery Agent Analytics Structured agent events such as LLM calls and tool use ADK Python with the plugin Opt-in with --bq-analytics
Third-party integrations Telemetry in external platforms Any OpenTelemetry-instrumented agent Opt-in per provider

Cloud Trace

For ADK, spans nest as invoke_workflow, then invoke_agent, then call_llm (with generate_content) and execute_tool. Setup is automatic for Agent Runtime and scaffolded Cloud Run or GKE projects; manual deployments configure their own exporter. You view traces in the Cloud Console under Trace explorer. Google's Cloud Trace page in the agents-cli docs covers the same setup.

Prompt-response logging

This tier needs Terraform-provisioned resources: a service account, a GCS bucket and a BigQuery dataset. The skill says to run agents-cli infra single-project --apply --project PROJECT_ID to create them, before agents-cli deploy.

Content capture has two independent tiers. GCS and BigQuery completions capture full prompts and responses when OTEL_INSTRUMENTATION_GENAI_COMPLETION_HOOK=upload and LOGS_BUCKET_NAME are set. Trace spans and log events are governed by OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT, which defaults to NO_CONTENT. In the skill's words, the net Terraform-deploy default is "full content in GCS/BigQuery, none in traces."

BigQuery Agent Analytics and feedback

The optional ADK plugin writes structured events to BigQuery for conversational analytics, LLM-as-judge evals, Looker Studio dashboards and tool provenance. A reference file also shows how to add a user-feedback endpoint whose structured logs flow through a log sink into BigQuery.

Third-party platforms

A comparison table covers AgentOps, Arize AX, Phoenix, MLflow, Monocle, Weave and Freeplay, with setup complexity and whether each can be self-hosted. The skill points to the ADK integrations directory for the full list.

Four reference files support it: cloud-trace-and-logging.md, bigquery-agent-analytics.md, feedback-mechanism.md and adk-docs.md.

How do you install it?

The Getting Started guide lists Python 3.11+, uv and Node.js as prerequisites, with the Google Cloud SDK and Terraform as optional extras for deployment. The recommended install adds the CLI and all seven skills:

uvx google-agents-cli setup

Or install skills only, either the whole set or just this one (the command shown on skills.sh):

npx skills add google/agents-cli
npx skills add https://github.com/google/agents-cli --skill google-agents-cli-observability

The skill's metadata lists the agents-cli binary as a requirement, installed with uv tool install google-agents-cli. It assumes a project scaffolded with agents-cli. In Claude Code, run /skills to confirm the agents-cli skills are listed. Codex and Cursor read the same SKILL.md format; see Codex skills and Cursor skills.

How do you use it?

The skill loads when you ask to "set up tracing", "monitor my agent", "configure logging", "add observability" or "debug production traffic", per its description. You can also call /google-agents-cli-observability in Claude Code.

Example prompts (suggestions, not quotes from the repo):

  • Set up observability for this agent before the first deploy. Ask me which tiers I need.
  • No traces show up in Cloud Trace for the Cloud Run deployment. Check the likely causes.
  • Keep prompt content out of traces but keep full completions in BigQuery. What do I set?
  • Add a thumbs up and thumbs down feedback endpoint and route it to BigQuery.
  • Compare Phoenix and MLflow for this agent. We need a self-hosted option.

Tips and limits

From the SKILL.md, the repo docs and skills.sh, not from testing:

  • Order matters. For Terraform-managed observability, run infra single-project --apply first, then deploy. If you already deployed imperatively, the skill says: "Do not apply Terraform afterward."
  • Full content is stored by default. Completions in GCS and BigQuery capture full prompts and responses and ignore the content-capture variable. To stop that, remove the upload variables.
  • Python and Go differ. ADK Python reads the content variable as an enum (NO_CONTENT, EVENT_ONLY, SPAN_ONLY, SPAN_AND_EVENT) and treats true or false as invalid. ADK Go reads it as a boolean.
  • Some platforms replace native telemetry. The skill notes this for AgentOps.
  • Costs. For high telemetry costs, it suggests turning content capture off, reducing BigQuery retention and disabling unused tiers. Resources are created in your own Google Cloud project.
  • Security audits. On skills.sh, Gen Agent Trust Hub rated it Pass with risk level Safe (22 October 2026); its report flags free-text user feedback logged to BigQuery as an ingestion point for untrusted data and lists PII redaction or table expiration as mitigations. Socket (20 October 2026) and Snyk (risk level Low, 20 October 2026) rated it Pass.

Under Apache 2.0 you can copy and modify the skill as long as you keep the license and notices and mark files you change.

Save it in your Kipt library

In Kipt, open Skills and paste the skill's GitHub folder link to import it with its four reference files, or paste the repo link and pick from the list of all seven agents-cli skills. You can also drop the folder or a .zip. Your skills live in your own private GitHub repo and install into Claude Code or Codex with one command:

npx skills add https://github.com/<you>/<repo>.git --all -g

Kipt is not affiliated with Google.

Similar skills

Questions

What is the google-agents-cli-observability skill?

It is one of seven Agent Skills in Google's google/agents-cli repo. It teaches a coding agent to monitor agents deployed with agents-cli, using four tiers: Cloud Trace, prompt-response logging, the BigQuery Agent Analytics plugin and third-party platforms such as AgentOps, Phoenix and MLflow. It also has a troubleshooting table.

Is Cloud Trace on by default for agents-cli projects?

Yes. The skill says Cloud Trace works out of the box with no infrastructure needed, using OpenTelemetry spans wired at app startup. On Agent Runtime it is gated on GOOGLE_CLOUD_AGENT_ENGINE_ENABLE_TELEMETRY, which deploy sets. You view traces in the Cloud Console under Trace explorer.

Does prompt-response logging store full prompts?

By default after a Terraform deploy, yes in GCS and BigQuery, and no in traces. The skill describes two independent tiers: completions uploaded to GCS and BigQuery capture full content, while trace spans and Cloud Logging events default to NO_CONTENT. Removing the upload variables stops the full-content capture.

How do I install the agents-cli observability skill?

Run uvx google-agents-cli setup to install the agents-cli command and all seven skills, or add only this one with npx skills add https://github.com/google/agents-cli --skill google-agents-cli-observability. Logging and analytics need Terraform-provisioned infrastructure in a Google Cloud project.

Does the observability skill work with ADK Go agents?

Partly. The skill says Go projects get Cloud Trace and the BigQuery telemetry stack, but the GCS completion upload behind prompt-response logging and the BigQuery Agent Analytics plugin are Python only. Go also reads the content-capture variable as a boolean rather than an enum.

I already deployed with agents-cli deploy. Can I add Terraform observability later?

The skill says not to apply Terraform afterward, because on Agent Runtime Terraform owns the whole engine. Either delete the deployment and start over, or keep it, set the observability environment variables with agents-cli deploy --update-env-vars and grant the service account the telemetry IAM roles yourself.

More in Skill directory

All guides

Keep your prompts and skills in one place, in your own GitHub. Install every skill with one command.

Get started

Kipt is an independent product, not affiliated with or endorsed by Anthropic, OpenAI, Cursor or GitHub. Claude, ChatGPT, Codex, Cursor and GitHub are trademarks of their owners.