@siriusai_cybernetics/stacksy
v0.3.0
Published
Onboarding CLI for Stacksy Discovery Lab specialists. Installs the Claude Desktop bundle + Knowledge file in one command.
Maintainers
Readme
@siriusai_cybernetics/stacksy — Stacksy specialist onboarding CLI
Tier 2.5 npm-publishable package (2026-05-29). Operator-facing distribution guide. Specialists see
docs/runbooks/SPECIALIST_ONBOARDING_2026-05-29.md§A.Distributed under the
@siriusai_cyberneticsnpm org (operator-controlled via siriusai.nl). The CLI + product stay brandedstacksy— parent-brand (Siriusai) / product-brand (Stacksy) split.
What it does
One command takes a specialist from "nothing installed" to "ready to fire first Interpolis run in Claude Desktop" in ~2 minutes.
npx @siriusai_cybernetics/stacksy@latest onboardOr, after global install (npm install -g @siriusai_cybernetics/stacksy):
stacksy onboardPackage layout
scripts/onboard/
├── package.json @siriusai_cybernetics/stacksy npm package metadata
├── README.md this file
├── bin/
│ └── stacksy.mjs bin shim — routes `stacksy <command>`
└── lib/
└── onboard.mjs the `stacksy onboard` command implementationCLI options
stacksy onboard [options]
--env <prod|staging> Skip the env prompt (default: ask)
--bundle-url <URL> Fetch the .mcpb from this URL
(default: resolve from repo-local
extensions/stacksy.mcpb)
--knowledge-url <URL> Fetch Knowledge file from this URL
(default: resolve from repo-local
docs/handoff/STACKSY_SPECIALIST_KNOWLEDGE.md)
--skip-open Don't auto-open the .mcpb
(specialist double-clicks manually)When called via npx @siriusai_cybernetics/stacksy@latest, the CLI is NOT inside a repo checkout — so --bundle-url and --knowledge-url MUST be set (typically baked into the operator's hosted entrypoint script).
Operator publish flow (~50 min total)
Step 1 — Register the @siriusai_cybernetics npm org (one-time, ~5 min)
Status (2026-05-29): DONE. Org registered. Skip to Step 2.
Visit https://www.npmjs.com/org/create and register siriusai_cybernetics as an organization. Add yourself ([email protected] or your npm account) as the owner.
- Free plan = public packages only
- Paid plan = restricted packages (specialists need to authenticate to install)
Recommended for the pilot: free plan + publishConfig.access: "public" so specialists can npx without credentials. Flip to restricted later once specialist count + monetization model justify the change.
If you go public, update package.json:
{
"publishConfig": {
"access": "public"
}
}Step 2 — Publish the package (~10 min including login)
cd scripts/onboard
npm login # uses your @siriusai_cybernetics org credentials
npm publish # publishes @siriusai_cybernetics/stacksy@<version>Verify on https://www.npmjs.com/package/@siriusai_cybernetics/stacksy.
Step 3 — Host the release artifacts (~30 min)
The CLI fetches the .mcpb and Knowledge file from URLs you control. Recommended setup using siriusai.nl + Cloud Storage:
Cloud Storage public bucket:
gsutil mb -p western-beanbag-445419-q5 -l europe-west4 gs://siriusai-stacksy-releases/
gsutil iam ch allUsers:objectViewer gs://siriusai-stacksy-releases/Auto-upload from the bundle build script — add to extensions/stacksy-mcpb/build.sh:
# After `mcpb pack`:
if [[ "${STACKSY_PUBLISH_RELEASE:-false}" == "true" ]]; then
gsutil cp extensions/stacksy.mcpb \
gs://siriusai-stacksy-releases/latest/stacksy.mcpb
gsutil cp docs/handoff/STACKSY_SPECIALIST_KNOWLEDGE.md \
gs://siriusai-stacksy-releases/latest/STACKSY_SPECIALIST_KNOWLEDGE.md
fiRun STACKSY_PUBLISH_RELEASE=true bash extensions/stacksy-mcpb/build.sh to publish.
URLs become:
https://storage.googleapis.com/siriusai-stacksy-releases/latest/stacksy.mcpbhttps://storage.googleapis.com/siriusai-stacksy-releases/latest/STACKSY_SPECIALIST_KNOWLEDGE.md
Step 4 — Host the install one-liner (DONE — via the release bucket)
Status (2026-05-29): DONE. The install script is hosted directly in the public Cloud Storage release bucket — no separate frontend hosting needed. URL:
https://storage.googleapis.com/siriusai-stacksy-releases/install
How it was set up (for reference / re-deploy):
# scripts/onboard/install-redirect.sh content uploaded to the bucket
gsutil cp scripts/onboard/install-redirect.sh gs://siriusai-stacksy-releases/install
gsutil setmeta -h "Content-Type:text/x-shellscript" gs://siriusai-stacksy-releases/install
# Windows PowerShell entrypoint (2026-06-11):
# irm https://storage.googleapis.com/siriusai-stacksy-releases/install.ps1 | iex
gsutil cp scripts/onboard/install-redirect.ps1 gs://siriusai-stacksy-releases/install.ps1
gsutil setmeta -h "Content-Type:text/plain" -h "Cache-Control: no-cache, max-age=0" gs://siriusai-stacksy-releases/install.ps1The script content:
#!/usr/bin/env bash
# https://storage.googleapis.com/siriusai-stacksy-releases/install
# Stacksy specialist onboarding entrypoint.
# Source: github.com/Hyperborean42/siriusai-stacksy (scripts/onboard/)
exec npx -y @siriusai_cybernetics/stacksy@latest onboard \
--bundle-url https://storage.googleapis.com/siriusai-stacksy-releases/latest/stacksy.mcpb \
--knowledge-url https://storage.googleapis.com/siriusai-stacksy-releases/latest/STACKSY_SPECIALIST_KNOWLEDGE.mdSpecialist final shape (live):
curl -sSL https://storage.googleapis.com/siriusai-stacksy-releases/install | bashStep 5 (optional polish) — Pretty URL via siriusai.nl/install/stacksy
Optional follow-up to shorten the install URL. The Cloud Storage URL works fine for the pilot; this is purely cosmetic.
Two viable paths whenever you have time:
- Cloud Run subdomain redirect — add an
install.siriusai.nlCloud Run domain mapping + tiny redirect service. Then specialists runcurl -sSL https://install.siriusai.nl | bash. ~30 min operator-side. - TransIP/DNS subpath — host the same bash script at
https://siriusai.nl/install/stacksyvia your frontend's static file config. Path depends on your frontend stack.
Until then: the Cloud Storage URL is the canonical install URL — stable, signed, public, free.
Pre-publish testing (locally)
The CLI works inside a repo checkout without needing npm publish:
# From repo root:
node scripts/onboard/bin/stacksy.mjs onboard
# Or with explicit overrides:
node scripts/onboard/bin/stacksy.mjs onboard --env stagingThis is the path you'd use for pilot specialists who have repo access while npm publish is pending.
To test the npm-installed shape:
cd scripts/onboard
npm pack # creates siriusai_cybernetics-stacksy-<version>.tgz
npm install -g ./siriusai_cybernetics-stacksy-*.tgz
stacksy --version # should print 0.1.0
stacksy onboard --env prod --bundle-url <test-bundle-URL>What's deferred to Tier 2.5+ / Tier 3
| Item | Status | Notes |
|---|---|---|
| Browser OAuth provisioning | Deferred | Needs admin panel + backend self-service endpoint; ~1-2 days. Spec §C.2 |
| @siriusai_cybernetics npm org registration | Operator-action | 5 min on npmjs.com; gates everything else |
| Cloud Storage release hosting | Operator-action | Configure gsutil cp in build pipeline; ~30 min |
| https://siriusai.nl/install/stacksy pretty URL (optional) | Operator-action | Cosmetic only — the GCS URL works today. Path depends on frontend stack (~30 min if Cloud Run subdomain; varies for Vercel/TransIP) |
| Claude Desktop Project API integration | Researching | API availability uncertain; defer |
| Windows CLI support + install.ps1 one-liner | DONE (2026-06-11) | Clipboard (clip.exe), auto-open (cmd /c start), PowerShell entrypoint. Needs Node 18+ like macOS |
| Windows installer (.msi) | Deferred | Would only remove the Node.js prerequisite; revisit after 5 specialists onboarded |
| Auto-update mechanism | Deferred | Tier 3 D.1 |
Failure recovery
- "Bundle not found" — provide
--bundle-urlOR run from inside a repo checkout - "Knowledge file not found" — provide
--knowledge-urlOR run from repo (Knowledge install is optional) - "Couldn't auto-open the bundle" — Claude Desktop not installed or the .mcpb file association is broken; specialist double-clicks the .mcpb manually (CLI prints the path)
- 37 tools don't show up post-install — bundle is stale; rebuild via
bash extensions/stacksy-mcpb/build.sh+ re-host + re-run CLI
Source
- Package:
scripts/onboard/(this directory) - Spec:
docs/specs/SPECIALIST_UX_HARDENING_2026-05-29.md§C (Tier 2) - Specialist runbook:
docs/runbooks/SPECIALIST_ONBOARDING_2026-05-29.md - Knowledge file pointer:
docs/handoff/STACKSY_SPECIALIST_KNOWLEDGE.md§I
