Install once, and your coding agent gains a full Google Cloud delivery team — a vendor-objective solution designer, a GCP architect, an agentic-systems architect, plus security, SRE, and QA — wired into a design-first, security-first delivery gate.
All the GCP skills. All the MCP servers available today. Assembled to fast-track the right architecture for your requirements.
Skill libraries for Google Cloud are everywhere. What they miss is the operating contract: who designs before anyone writes code, who signs off on least-privilege before anything deploys, what "done" means when an agent ships to a billable cloud. This plugin ships that contract, with the GCP know-how built in.
One install gives an agent (Claude, Codex, or Antigravity) six personas, eleven service skills, and a research pipeline that keeps every reference current. A plugin that goes stale in silence is worse than none.
The plugin is GCP-only and opinionated by design. It covers the services a builder reaches for — Firebase, Cloud Run, BigQuery, GKE — and the agent frameworks now built on top: ADK, and the MCP / A2A / AP2 protocol stack, deployed on the Gemini Enterprise Agent Platform. Refine it for your team. Keep the contract coherent as you do.
Speed is easy. Safe speed is engineered.
One plugin, four coding assistants. Skills live in skills/; each harness reads its own manifest.
Claude Code
/plugin marketplace add jpantsjoha/googlecloud-plugin
/plugin install googlecloud-plugin@googlecloud-plugin-marketplace
Antigravity (Gemini)
agy plugin install https://github.com/jpantsjoha/googlecloud-pluginKimi Code
/plugins install https://github.com/jpantsjoha/googlecloud-plugin
Codex — no install step. Codex discovers .agents/skills/ natively and reads AGENTS.md as its always-on adapter. Clone the repo and Codex picks up the contract:
git clone https://github.com/jpantsjoha/googlecloud-pluginAfter cloning to work on the plugin itself, install the pre-commit gate once:
make hooks # runs `make gate` (validate + manifest + mermaid + lint + test) on every commitThe plugin running in Antigravity — asked what it is and how it delivers value, the agent introduces the team and the delivery gate:
Every piece of work flows through sequential gates. No gate opens until the one before it clears. Human intent enters at the top; a human authorises production at the bottom. The AI personas scale the execution in between — they never grant their own authority.
%%{init: {'theme':'neutral', 'sequence': {'useMaxWidth': true, 'diagramMarginX': 24, 'diagramMarginY': 18, 'actorMargin': 44, 'boxMargin': 15, 'boxTextMargin': 8, 'noteMargin': 12, 'messageMargin': 42}}}%%
sequenceDiagram
autonumber
actor Req as Requirement
participant Sol as solution-designer
participant Arch as gcp-architect +<br/>agent-architect
participant Sec as gcp-security
participant Impl as Service Skills
participant QA as gcp-qa
actor Ops as gcp-ops
rect rgb(238, 242, 255)
note over Req,Sol: Solution Gate — is this even GCP? (skip if unambiguously GCP-only)
Req->>Sol: 1. Submit intent / problem
Sol-->>Arch: 2. Master HLD + vendor scope (hands GCP slice down)
end
rect rgb(237, 233, 254)
note over Arch: GCP Design Gate — design before code
Arch->>Arch: 3. GCP HLD/LLD + ADRs (agentic: topology, protocols, model, eval set)
end
rect rgb(252, 231, 243)
note over Arch,Sec: Security Gate — reviews the design, not the diff
Arch->>Sec: 4. Submit design
Sec-->>Impl: 5. Least-privilege IAM, no secrets, WAF pillar — pass or BLOCK
end
rect rgb(240, 253, 244)
note over Sec,Impl: Implementation — known-good GCP patterns, cost-warned
Impl->>Impl: 6. Build on google-cloud patterns, warn before every billable action
end
rect rgb(254, 249, 231)
note over Impl,QA: Quality Gate — receipts, not polish
Impl->>QA: 7. Submit candidate
QA-->>Ops: 8. make gate (all checks) + agent-eval — pass or BLOCK
end
rect rgb(224, 242, 254)
note over QA,Ops: Operational Readiness Gate — no dark production
Ops->>Ops: 9. SLO + alerting + runbook present?
Ops-->>Req: 10. Authorise release & emit status
end
| Gate | Owner | Blocks release when |
|---|---|---|
| Solution | solution-designer |
Vendor choice unclear, or GCP scope not delimited (skips if GCP-only) |
| GCP Design | gcp-architect (+ agent-architect if agentic) |
No GCP HLD, or HLD contradicts the solution design |
| Security | gcp-security |
IAM over-permissive, secrets in code, GCP risks unaddressed |
| Quality | gcp-qa (+ agent-architect agent-eval) |
make gate fails, acceptance criteria unmet, agent eval fails |
| Operational Readiness | gcp-ops |
No SLO, alerting, or runbook for a production workload |
The rule underneath every gate: infer intent, never infer permission. An AI persona drafts, designs, and validates; a named human authorises anything consequential, billable, or irreversible.
Six personas — the team:
| Persona | Tier | Owns |
|---|---|---|
solution-designer |
1 · vendor-agnostic | The overarching solution; objective vendor selection across GCP / AWS / Azure |
gcp-architect |
2 · GCP | GCP infrastructure design — HLD/LLD, ADRs, IAM and network topology |
agent-architect |
2 · GCP agentic | Agentic systems on GEAP — ADK, Agent Runtime, MCP/A2A/AP2, and agent evaluation |
gcp-security |
3 | Least-privilege IAM, secrets, Well-Architected security pillar |
gcp-ops |
3 | SLOs, alerting, runbooks, operational readiness |
gcp-qa |
3 | Validation, linting, smoke tests, the quality gate |
Eight first-wave service skills: Cloud Run · GKE · IAM · BigQuery · Cloud Storage · Vertex AI · Networking · Logging/Monitoring.
Three cross-cutting skills: mcp-servers (Google-managed + self-hosted MCP setup) · well-architected (the six-pillar framework) · terraform-gcp (CFT + provider patterns).
The plugin validates itself. Every skill conforms to a machine-readable contract; every reference URL is checked live; every source carries a retrieval date and content hash for audit.
make gate # the 3-minute gate: validate + lint + test
make validate # SKILL.md frontmatter contract
make lint # every reference URL resolves (HTTP 200)
make test # skill smoke tests
make check # freshness: content-hash drift vs live GCP docs
make crawl # refresh the research corpus (weekly)Freshness is a feature, not a chore: make check runs weekly and flags any GCP doc that has drifted since it was last captured, so the plugin never quietly rots.
Created and maintained by Jaroslav Pantsjoha (JP) — Technical Director and Enterprise Agent Solution Architect at Cognizant, Google Developer Expert (Google Cloud), speaker, and author of Building the Agentic Enterprise on Google Cloud (Packt).
I built this to make my own Google Cloud work repeatable, then to share it. The model is the easy part; the durable engineering is the harness around it — the skills, rules, MCP servers, gates, and evals. GCP patterns outlast the week's model release, so they are worth encoding once and reusing. This plugin is that baseline aimed squarely at Google Cloud — the GCP companion to my join-the-team harness. Part of the #HarnessEngineering body of work — the engineering discipline behind the Agentic Enterprise.
- LinkedIn: uk.linkedin.com/in/johas
- Google Developer Expert: me.developers.google.com/u/jpantsjoha
Open source under the MIT License — free for personal and commercial use, modification, and redistribution.
This is an independent, community project. It is not affiliated with, sponsored by, or endorsed by Google LLC. "Google Cloud", "GCP", "Gemini", "Vertex AI", and related names are trademarks of Google LLC, used here only to describe what the plugin works with.