Releases: jpantsjoha/googlecloud-plugin
Release list
v0.2.0 — Agent Plugins 1.0.0 & Agent Skills conformance
Standards release. googlecloud-plugin now conforms to Agent Plugins 1.0.0 and the Agent Skills specification — the open, vendor-neutral packaging standards co-maintained by Amazon, Cursor, Google, Microsoft, OpenAI and Vercel (announcement).
Why this matters
Before this release, a conformant Agent Plugins client would have loaded nothing from this plugin. Two independent fatal faults:
plugin.jsonomitted the required$schema. The manifest schema is closed, so §5.2 makes the omission fatal — the client rejects the plugin outright.- All 17 skills carried top-level frontmatter outside the closed Agent Skills field set. §6.1 requires clients to skip every non-conforming skill.
Standards adhered to
| Standard | Version | Scope | Enforced by |
|---|---|---|---|
| Agent Plugins | 1.0.0 | plugin.json, mcp.json, skills/ layout |
make spec |
plugin.schema.json |
1.0.0 | root manifest, closed field set | make spec |
mcp.schema.json |
1.0.0 | MCP manifest, transport discriminator | make spec |
| Agent Skills | current | all 17 SKILL.md |
make validate |
| Model Context Protocol | stdio, 2025-06-18 | mcp.json servers |
live handshake |
Breaking — skill authors and downstream tooling
Frontmatter is now {name, description, license, metadata}. The top-level version, triggers, required_scopes, mcp_servers, persona, tier, gate and owns_eval fields are gone; the same data lives under metadata, namespaced, as string values:
metadata:
"googlecloud-plugin/required-scopes": "run.services.create, run.services.get"Split on ", ". Routing keywords moved into description — the only field a conformant client routes on.
No action for users. Existing installs are unaffected; the per-harness manifests are unchanged.
Verification
| Check | Result |
|---|---|
skills-ref — official Agent Skills reference validator |
17/17 valid |
check-jsonschema vs both published schemas |
pass / pass |
agy plugin validate — Antigravity loader |
17 skills, 2 mcpServers, exit 0 |
| Antigravity routing | 17 visible, 6/6 correct |
| Kimi routing | 17 loaded, 7/7 correct |
| Live MCP handshake | gcloud-mcp-server v0.5.3, run_gcloud_command |
make gate |
spec + validate + manifest + mermaid + 102 URLs + 195 tests |
Verify it yourself, without trusting this repo's scripts:
uvx --from "git+https://github.com/agentskills/agentskills.git#subdirectory=skills-ref" skills-ref validate skills/cloud-run
uvx check-jsonschema --schemafile https://agent-plugins.org/schemas/1.0.0/plugin.schema.json plugin.jsonNot verified: Codex (usage limit at release time) and the Toolbox MCP server (opt-in by design — it needs a user-specific tools.yaml).
Also in this release
- New
make specconformance gate; tests 136 → 195 - Eight skill descriptions gained boundary clauses after cross-harness routing tests found four ambiguous pairs
mcp.jsondeclares noenv— only\${PLUGIN_ROOT}/\${PLUGIN_DATA}expand, so\${GCP_PROJECT_ID}would have been passed through literally. Export it in your shell instead.- ADR-007 · CHANGELOG
Full changelog: v0.1.0...v0.2.0
googlecloud-plugin v0.1.0 — a Google Cloud delivery team for your coding agent
Human-led. Agent-powered. Built for Google Cloud.
Install once, and your coding agent gains a full Google Cloud delivery team — not just a pile of skills, but the operating contract that makes them safe to use on a real, billable cloud.
What it is
A GCP-only, opinionated plugin that installs six personas and wires them into a sequential, design-first, security-first gate:
- solution-designer — vendor-objective; decides if GCP is even the right home before anyone builds
- gcp-architect — GCP infrastructure design, HLD/LLD, ADRs
- agent-architect — agentic systems on the Gemini Enterprise Agent Platform: ADK, Agent Runtime, and the MCP / A2A / AP2 protocol stack, plus agent evaluation
- gcp-security — least-privilege IAM, secrets, Well-Architected security, reviewed at design time
- gcp-ops — SLOs, alerting, runbooks, operational readiness
- gcp-qa — validation, linting, smoke tests
Plus eight service skills (Cloud Run, GKE, IAM, BigQuery, Cloud Storage, Vertex AI, networking, observability), MCP server setup, and a research pipeline that keeps every reference current.
How it fast-tracks value
- Design before code, security before deploy. The gate stops the expensive mistakes — over-permissioned IAM, public buckets, ungoverned spend — before they reach production.
- Install once, run anywhere. Validated on Claude Code, Antigravity, Codex, and Kimi.
- Infer, don't interrogate.
make preflightreads your project and credentials from the environment and proceeds; it prompts only when something is genuinely missing. - Never silently stale. A weekly freshness job re-checks every GCP source and flags drift, so guidance stays current as the platform moves.
- Receipts, not vibes.
make gate— contracts, packaging, diagram, links, tests — runs on every commit and in CI.
Start with make preflight, then let the gate carry the work from intent to release.
This is an independent community project, not affiliated with or endorsed by Google LLC.
— JP, your friendly GDE