Skip to content
Agentmart
Submit a tool

start

Triage entry point for multi-cloud (AWS / Azure / GCP) infrastructure work across local / test / staging / production: resolves the OKF k8s/tf bundle, fixes…

Skill by wfukatsu

IndexedMIT★ 13
Write a review

Chalk Labs

Need help deploying start?

Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.

Book a call

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/wfukatsu-start

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Triage entry point for multi-cloud (AWS / Azure / GCP) infrastructure work across local / test / staging / production: resolves the OKF k8s/tf bundle, fixes environment and cloud, then routes to design, implement or review. Use for any infrastructure request that does not yet name a mode.

Use cases: design, testing, devops, cloud

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md and code. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/start

What's inside

Skill name
start
Description
Triage entry point for multi-cloud (AWS / Azure / GCP) infrastructure work across local / test / staging / production: resolves the OKF k8s/tf bundle, fixes environment and cloud, then routes to design, implement or review. Use for any infrastructure request that does not yet name a mode.
License
MIT
Instructions
179 lines in SKILL.md · view on GitHub

Agent Skills spec issues

  • name: missing
  • allowed-tools: must be a space-separated string

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Infrastructure Triage and Routing

Shared files: a path written `@rules/…`, `@skills/…`, `@templates/…` or `@docs/…` is relative to the
plugin root, `${CLAUDE_PLUGIN_ROOT}` — not to the project being worked on. Read it from there.

**Usage:** `/infra:start [target] [--env=<env>] [--cloud=<cloud>]`

Ground every infrastructure decision in the OKF `okf-k8s-tf` bundle
(@rules/okf-k8s-tf-bundle.md) rather than in model memory, and hand the real work to the skill
that is sized for it.

## Desired Outcome

Four things are settled and passed downstream, so the mode skill never re-asks them:

1. The resolved bundle root `$OKF`, and its **freshness list**: the documents past `stale_after` and
   those awaiting re-verification, each with its reason
2. The **target environment(s)** — `local` / `test` / `staging` / `production`
3. The **target cloud(s)** — AWS / Azure / GCP, and which environments exist in each
4. The **mode** — design, implement or review — and the target path

## Decision Criteria

| The user wants | Mode | Route to |
|----------------|------|----------|
| A configuration decided from requirements; technology selection; an approach | **design** | `/infra:design` |
| Terraform / manifests / values / CI written or fixed | **implement** | `/infra:implement` |
| Existing code or a design document assessed, with findings | **review** | `/infra:review` |

Ambiguous? Ask **one** question with AskUserQuestion. A compound request ("design it and then
build it") splits into design → user confirmation → implement; never proceed to implementation
without agreement on the design.

**Route by invoking the mode skill with the Skill tool**, passing the four settled facts as
arguments. This skill does triage only — design and review carry a higher reasoning tier
(see Model Policy).

The exception is a single factual lookup ("what does the bundle say about Vault HA?") or a
request whose whole point is the triage itself. Answer those here.

## Prerequisites

| Input | Required/Recommended | Source |
|-------|---------------------|--------|
| `knowledge/okf-k8s-tf/` (the bundle) | **Required** | Vendored; `tools/update-okf-bundle.sh status --bundle=k8s-tf` |
| The target repository or design document | Recommended | The user |
| reports/03_design/target-architecture.md | Recommended | /architect:design-microservices |
| reports/08_infrastructure/infrastructure-design.md | Recommended | /architect:design-infrastructure |

## Step 1 — Resolve the bundle

Run `${CLAUDE_PLUGIN_ROOT}/tools/update-okf-bundle.sh status --bundle=k8s-tf` and take the
resolved path as `$OKF`. The resolution order, where its upstream comes from, and what to do when
it cannot be found are all in @rules/okf-k8s-tf-bundle.md §1.

**If the bundle cannot be resolved, do not answer from memory.** It is the primary source for
this skill, not a convenience.

## Step 2 — Check freshness

The same `status` output ends with the **freshness list**, in two parts:

- `past stale_after` — computed for today from each document's frontmatter.
- `awaiting re-verification` — the list the weekly upstream refresh recorded, with the reasons:
  a cited page changed, redirected to a different page or stopped answering, a stated version fell
  further behind or reached end of life. It is dated (`as of`); a list older than the latest
  Monday means the refresh has not run since.

Carry both parts forward unchanged as the freshness list. A listed document is still citable,
as dated evidence: every claim resting on it says it awaits re-verification and why, and when the
claim matters, check the locally held page (`knowledge/okf-k8s-tf-upstream/pages/`) or the
official source (@rules/okf-k8s-tf-bundle.md §1, §6). Revising the document is not this skill's
job — tell the user `/architect:revise-knowledge` does it.

## Step 3 — Fix the environment

Infrastructure has four environments and **no output is produced before the environment is
settled**. Use `--env=<env>`; when absent, ask with AskUserQuestion. When several are in scope,
the deliverable gets one section per environment.

Coverage differs sharply per environment, and the difference is not cosmetic:

| Environment | Bundle | Posture in the output |
|-------------|--------|-----------------------|
| `local` | **Absent** | Say "outside the bundle's scope". Do not assert |
| `test` | Observed implementation (a procedure or CI `kubectl apply`, direct apply) | Citable as fact |
| `staging` | Observed implementation (Argo CD app-of-apps, GitOps) | Citable as fact |
| `production` | **No observed implementation** — guidance only | Say the investigated repositories contain no production implementation, and present production as staging plus approval/protection/sync windows |

Full matrix and promotion path: @rules/infra/environments.md.

## Step 4 — Fix the cloud

Multi-cloud is the default premise; never answer as if one cloud were assumed. Use
`--cloud=<cloud>`; when absent, ask. Settle the four items in @rules/infra/multi-cloud.md §0 and
fill the cloud × environment grid before routing — an empty cell means "not deployed here", and
must not be left implicit.

## Step 5 — Route

Pick the mode from Decision Criteria and invoke the corresponding skill, passing environment,
cloud, target path, `$OKF` and the freshness list.

## Non-Negotiable Rules

These bind the mode skills too.

1. **Cite the bundle for anything it covers.** Never write from memory. Attach
   `[foundation/terraform.md]`-style sources to each claim (@rules/okf-k8s-tf-bundle.md §7). A
   claim resting on a document on the freshness list says the document awaits re-verification,
   and why.
2. **Say "outside the bundle's scope" when it is.** Then WebFetch official documentation, or
   leave it as an open question. Do not assert.
3. **Keep fact, guidance and open question separate.** The bundle itself distinguishes observed
   implementation / design guidance / open question; the output preserves the same three tiers
   (@rules/okf-k8s-tf-bundle.md §4).
4. **Multi-cloud is the default.** Pass every conclusion through the portability boundary in
   @rules/infra/multi-cloud.md before writing it.
5. **There are four environments.** Differences stay in overlay and values as **value

From wfukatsu/nexus-architect, MIT.

More from wfukatsu/nexus-architect

Signals

Stars
13
Forks
1
Open issues
0
Last commit (this folder)
Oct 8, 2026 (28 days ago)
Last release
Oct 9, 2026 (27 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from wfukatsu

Done for you

Rather not set it up yourself?

Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.

See how we build AI agents →Read the case studies

Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.

launches shipped
50+
launches shipped
media placements
500+
media placements
founders on every account
2
founders on every account

Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block

Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.

Get start running in your business

Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.

Goes to the Chalk Labs founders, not the newsletter. Privacy

Need help deploying start?

CallGet help
Cite this page

start skill. Agent Mart. https://agentmart.dev/skills/wfukatsu-start (last updated 2026-10-24)