google-agents-cli-workflow（Google Agent CLI 工作流）

ADK Development Workflow & Guidelines

STOP — Do NOT write code yet. If no project exists, scaffold first with agents-cli scaffold create <name>. If the user already has code, use agents-cli scaffold enhance. to add the agents-cli structure. Run agents-cli info to check if a project already exists. Skipping this leads to missing eval boilerplate, CI/CD config, and project conventions.

agents-cli is a CLI and skills toolkit for building, evaluating, and deploying agents on Google Cloud using the Agent Development Kit (ADK). It works with any coding agent — Gemini CLI, Claude Code, Codex, or others. Install with uvx google-agents-cli setup.

Requires: google-agents-cli = 0.1.2 If version is behind, run: uv tool install "google-agents-cli=0.1.2" Check version: agents-cli info Install uv first if needed.

Session Continuity & Skill Cross-References

Re-read the relevant skill before each phase — not after you've already started and hit a problem. Context compaction may have dropped earlier skill content. If skills are not available, run uvx google-agents-cli setup to install them.

Setup

If agents-cli is not installed:

uv tool install google-agents-cli

uv command not found

Install uv following the official installation guide.

Product name mapping

The platform formerly known as "Vertex AI" is now Gemini Enterprise Agent Platform (short: Agent Platform). Users may refer to products by different names. Map them to the correct CLI values:

The vertexai Python SDK package name is unchanged.

Phase 0: Understand

Before writing or scaffolding anything, understand what you're building.

If DESIGN_SPEC.md already exists, read it — it is your primary source of truth. Otherwise:

Do NOT proceed to planning, scaffolding, or coding. Ask the user the questions below and wait for their answers. You MUST have the user's answers before moving on. Do not assume, research, or fill in the blanks yourself. The user's intent drives everything — skipping this step leads to wasted work.

Always ask:

1. What problem will the agent solve? — Core purpose and capabilities
2. External APIs or data sources needed? — Tools, integrations, auth requirements
3. Safety constraints? — What the agent must NOT do, guardrails
4. Deployment preference? — Prototype first (recommended) or full deployment? If deploying: Agent Runtime, Cloud Run, or GKE?

Ask based on context:

• If retrieval or search over data mentioned (RAG, semantic search, vector search, embeddings, similarity search, data ingestion) → Datastore? Options: agent_platform_vector_search (embeddings, similarity search) or agent_platform_search (document search, search engine).
• If agent should be available to other agents → A2A protocol? Enables the agent as an A2A-compatible service.
• If full deployment chosen → CI/CD runner? GitHub Actions (default) or Google Cloud Build?
• If agent should remember user preferences or facts across sessions → Memory Bank? Long-term memory across conversations. See /google-agents-cli-adk-code.
• If Cloud Run or GKE chosen → Session storage? In-memory (default), Cloud SQL (persistent), or Agent Platform Sessions (managed).
• If deployment with CI/CD chosen → Git repository? Does one already exist, or should one be created? If creating, public or private?

Once you have the user's answers, write a DESIGN_SPEC.md with the user's approval. See /google-agents-cli-scaffold for how these choices map to CLI flags. At minimum include these sections — expand with more detail if the user wants a thorough spec:

# DESIGN_SPEC.md

## Overview
Describe the agent's purpose and how it works.

## Example Use Cases
Concrete examples with expected inputs and outputs.

## Tools Required
Each tool with its purpose, API details, and authentication needs.

## Constraints & Safety Rules
Specific rules — not just generic statements.

## Success Criteria
Measurable outcomes for evaluation.

## Reference Samples
Check the Notable Samples in Phase 1 — list any that match this use case.

Optional sections for more detailed specs: Edge Cases to Handle, Architecture & Sub-Agents, Data Sources & Auth, Non-Functional Requirements.

Once you have a clear understanding, proceed to Phase 1.

Phase 1: Study Reference Samples

Ask yourself: is there a sample that can help me design this and cut time? Scan the keywords below. Multiple samples can match — clone and study all that are relevant.

# Clone a sample to study — read the key files, understand the patterns, then apply
# them to your own scaffolded project. Do NOT use `adk@<sample>` scaffolding.
git clone --filter=tree:0 --sparse https://github.com/google/adk-samples /tmp/adk-samples 2>/dev/null; \
cd /tmp/adk-samples && git sparse-checkout add python/agents/<sample-name>

• ambient-expense-agent — Agent that runs on a schedule or reacts to events, with no interactive user. Keywords: scheduled, cron, daily, pubsub, event-driven, alerts, email, ambient Key files: expense_agent/fast_api_app.py, expense_agent/agent.py, expense_agent/config.py, terraform/
• adk-ae-oauth — Agent with OAuth 2.0 user consent, deployed to Agent Runtime with Gemini Enterprise. Keywords: OAuth, authentication, user consent, Google Drive, Agent Runtime, Gemini Enterprise Key files: README.md, adk_ae_oauth/tools.py, adk_ae_oauth/auths.py
• genmedia-for-commerce — Full-stack agent with React UI, MCP tools, media/image handling, and Gemini Enterprise registration. Keywords: MCP, media, video generation, Veo, virtual try-on, retail, full-stack, React, Gemini Enterprise Key files: genmedia4commerce/agent.py, genmedia4commerce/agent_utils.py, genmedia4commerce/fast_api_app.py
• deep-search — Research agent that iterates until quality is met, with source citations. Keywords: research, citations, iterative, grounding, multi-agent, human-in-the-loop, web search, report Key files: app/agent.py, app/config.py
• safety-plugins — Reusable safety guardrails that plug into any agent runner. Keywords: safety, guardrails, model armor, filters Key files: safety_plugins/plugins/model_armor.py, safety_plugins/plugins/agent_as_a_judge.py, safety_plugins/main.py
• data-science — Agent that executes code in a managed sandbox for data analysis. Keywords: SQL, BigQuery, code execution, sandbox Key files: data_science/sub_agents/analytics/agent.py
• memory-bank — Conversational agent with cross-session memory via Memory Bank (Cloud Run and Agent Runtime). Keywords: memory, cross-session, recall, context, remember, Memory Bank Key files: app/agent.py, app/agent_runtime_app.py, app/fast_api_app.py

If no sample matches, proceed to Phase 2. But first — are you sure? Re-read the user's request and compare it against the keywords above. Skipping a matching sample means rebuilding patterns that already exist.

IMPORTANT — Exit criteria: After studying a sample, ask yourself: can I apply anything from this sample to help me deliver the design? Note what you'll reuse before moving on. Do NOT proceed until you've answered this.

This list is useful at any phase — revisit it when you hit deployment, publishing, or infrastructure questions. A sample's Terraform or registration pattern may be exactly what you need later.

Phase 2: Scaffold (if needed)

Use /google-agents-cli-scaffold to create a new project or import an existing one into the agents-cli format (adding deployment, CI/CD, infrastructure). It covers architecture choices (deployment target, agent type, session storage) and project creation or enhancement.

Skip this phase if the project was already created or enhanced by agents-cli — run agents-cli info from the project root to check.

Phase 3: Build and Implement

Implement the agent logic:

1. Write/modify code in the agent directory (check GEMINI.md / CLAUDE.md for directory name)
2. Quick smoke test: Use agents-cli run "your prompt" to verify the agent works after changes — this is the fastest way to check behavior without leaving the terminal
3. Iterate on the implementation based on user feedback

If the user asks for interactive testing, suggest agents-cli playground — it opens a web-based playground for manual conversation with the agent.

For ADK API patterns and code examples, use /google-agents-cli-adk-code.

NEVER write pytest tests that assert on LLM output content (e.g., checking for keywords in responses, verifying persona, validating tone). LLM outputs are non-deterministic — these tests are flaky by nature and belong in eval, not pytest. Use agents-cli run for quick checks and agents-cli eval run for systematic validation.

Phase 3.5: Provision Datastore (RAG projects only)

For agentic_rag projects, provision the datastore before testing: agents-cli infra datastore, then agents-cli data-ingestion. Use infra datastore — not infra single-project (same datastore provisioning but faster, skips unrelated Terraform).

Phase 4: Evaluate

This is the most important phase. Evaluation validates agent behavior end-to-end.

MANDATORY: Activate /google-agents-cli-eval before running evaluation. It contains the evalset schema, config format, and critical gotchas. Do NOT skip this.

Do NOT skip this phase. After building the agent, you MUST proceed to evaluation. Do NOT write pytest tests to validate agent behavior — that is what eval is for.

uv run pytest vs agents-cli eval run — know the difference:

• uv run pytest — Tests *code correctness*: imports work, functions return expected types, API contracts hold. Does NOT test whether the agent behaves well.
• agents-cli eval run — Tests *agent behavior*: response quality, tool usage, persona consistency, safety compliance. This is what validates your agent actually works.
• agents-cli run "prompt" — Quick one-off smoke test during development. Use this for fast iteration, not pytest.

NEVER write pytest tests that check LLM response content (e.g., asserting pirate keywords appear, checking if the agent mentions allergies). LLM outputs are non-deterministic. Use eval with LLM-as-judge criteria instead.

1. Start small: Begin with 1-2 sample eval cases, not a full suite
2. Run evaluations: agents-cli eval run
3. Discuss results with the user
4. Fix issues and iterate on the core cases first
5. Only after core cases pass, add edge cases and new scenarios
6. Repeat until quality thresholds are met

Expect 5-10+ iterations here.

Phase 5: Deploy

Once evaluation thresholds are met:

1. Check if the project has a deployment target configured — run agents-cli info to see current config
2. If the project is a prototype (no deployment target), add deployment support first: agents-cli scaffold enhance. --deployment-target <target> See /google-agents-cli-deploy for the deployment target decision matrix (Agent Runtime vs Cloud Run vs GKE).
3. Deploy when ready: agents-cli deploy

IMPORTANT: Never deploy without explicit human approval.

Phase 6: Publish (optional)

Not all agents require this — currently supporting Gemini Enterprise. See /google-agents-cli-publish for registration modes, flags, and troubleshooting.

Phase 7: Observe

After deploying, use observability tools to monitor agent behavior in production. See /google-agents-cli-observability for Cloud Trace, prompt-response logging, BigQuery Analytics, and third-party integrations.

Operational Guidelines for Coding Agents

Common Shortcuts to Resist

Agents routinely skip steps with plausible-sounding excuses. Recognize these and push back:

Principle 1: Code Preservation & Isolation

Code modifications require surgical precision — alter only the code segments directly targeted by the user's request and strictly preserve all surrounding and unrelated code.

Mandatory Pre-Execution Verification:

Before finalizing any code replacement, verify the following:

1. Target Identification: Clearly define the exact lines or expressions to change, based *solely* on the user's explicit instructions.
2. Preservation Check: Confirm that all code, configuration values (e.g., model, version, api_key), comments, and formatting *outside* the identified target remain identical.

Example:

• User Request: "Change the agent's instruction to be a recipe suggester."
• Incorrect (VIOLATION): root_agent = Agent(name="recipe_suggester", model="gemini-1.5-flash", # UNINTENDED - model was not requested to change instruction="You are a recipe suggester.")
• Correct (COMPLIANT): root_agent = Agent(name="recipe_suggester", # OK, related to new purpose model="gemini-flash-latest", # PRESERVED instruction="You are a recipe suggester." # OK, the direct target)

Principle 2: Execution Best Practices

• Model Selection — CRITICAL:

- NEVER change the model unless explicitly asked. - When creating NEW agents (not modifying existing), use the latest Gemini model. List available models to pick the newest one: # Use 'global' or any supported region (e.g. 'us-east1') uv run --with google-genai python -c " from google import genai client = genai.Client(vertexai=True, location='global') for m in client.models.list(): print(m.name) " - Do NOT use older models unless explicitly requested. For model docs, fetch https://adk.dev/agents/models/google-gemini/index.md. See also stable model versions.

• Running Python Commands:

- Always use uv to execute Python commands (e.g., uv run python script.py) - Run uv sync before executing scripts

• Breaking Infinite Loops:

- Stop immediately if you see the same error 3+ times in a row - RED FLAGS: Lock IDs incrementing, names appending v5→v6→v7, "I'll try one more time" repeatedly - State conflicts (Error 409): Use terraform import instead of retrying creation - When stuck: Run underlying commands directly (e.g., terraform CLI)

• Troubleshooting:

- Check /google-agents-cli-adk-code first — it covers most common patterns - Use WebFetch on URLs from the ADK docs index (curl https://adk.dev/llms.txt) for deep dives - When encountering persistent errors, a targeted web search often finds solutions faster - CLI command failures: run agents-cli <command> --help — the output ends with a Source: line pointing to the exact source file implementing that command. Read it to understand the logic and diagnose failures. Use agents-cli info to get the full CLI install path if you need to browse across multiple files.

Systematic Debugging

When something breaks, follow this sequence — don't skip steps or shotgun fixes:

1. Reproduce — Run the exact command that failed. Save the full error output. If you can't reproduce it, you can't fix it.
2. Localize — Narrow the cause: is it the agent code, a tool, the config, or the environment? Use agents-cli run "prompt" to isolate agent behavior from deployment issues.
3. Fix one thing — Change one variable at a time. If you change the instruction AND the tool AND the config simultaneously, you won't know what fixed it (or what broke something else).
4. Verify — Rerun the exact reproduction command. Don't assume the fix worked.
5. Guard — If it was a non-obvious bug, add an eval case to catch regressions.

Stop-the-line rule: If a change breaks something that was working, stop feature work and fix the regression first. Don't push forward hoping to circle back — regressions compound.

• Environment Variables:

- .env files and env var assignments (e.g., GOOGLE_CLOUD_PROJECT, GOOGLE_CLOUD_LOCATION) are typically required for the agent to function — never remove or modify them unless the user explicitly asks - If a .env file exists in the project root, treat it as essential configuration - For secrets and API keys, prefer GCP Secret Manager over plain .env entries — see /google-agents-cli-deploy for secret management guidance

Using a Temporary Scaffold as Reference

When you need specific infrastructure files (Terraform, CI/CD, Dockerfile) but don't want to modify the current project, use /google-agents-cli-scaffold to create a temporary project in /tmp/ and copy over what you need.

Reference Files

Development Commands

Setup & Skills

Scaffolding

Development

Evaluation

Deployment & Infrastructure

Project Info

Use agents-cli info to discover the CLI install path — this is where the CLI source code lives. Read files under that path to understand CLI internals, command implementations, or template logic. The command only shows project details when run inside a generated agent project (i.e., one with [tool.agents-cli] in pyproject.toml).

Authentication

[!NOTE] When using an API key to authenticate, the login command does not persist them automatically, it just aids in retrieving them and providing instructions on how they can be persisted.

Skills Version

Troubleshooting hint: If skills seem outdated or incomplete, reinstall: `` agents-cli setup --skip-auth `` Only do this when you suspect stale skills are causing problems.

Related Skills

• /google-agents-cli-scaffold — Project creation, requirements gathering, and enhancement
• /google-agents-cli-adk-code — ADK Python API quick reference and production sample agents
• /google-agents-cli-eval — Evaluation methodology, evalset schema, and the eval-fix loop
• /google-agents-cli-deploy — Deployment targets, CI/CD pipelines, and production workflows
• /google-agents-cli-publish — Gemini Enterprise registration
• /google-agents-cli-observability — Cloud Trace, logging, BigQuery Analytics, and third-party integrations