@mesh-tech/mesh-cli 0.18.2 → 0.20.0

package/skills/core/SKILL.md

@@ -90,15 +90,15 @@ Top-level commands (each `register*` is wired in `src/program.ts`, which `bin/me
 | Command | Subcommands | Does | Key flags / args |
 |---|---|---|---|
 | `mesh login <context>` | (also `mesh logout <context>`) | Zitadel OIDC PKCE or device-code auth; caches creds in `~/.config/mesh/`. Context like `mesh.dev`. Auto-discovers config via SSM then HTTPS `.well-known/mesh.json`. | `--status` (reports the cached Zitadel session, or — when there isn't one — the AWS identity the credential chain resolves, so an SSO profile or a key/secret pair passes too; non-zero only when neither works), `--device`, `--export` (print AWS `export`s for `eval`), `--role <arn>`, `--region <r>` |
-| `mesh registry` | `login [context]`, `status`, `publish` | CodeArtifact auth → refreshes `~/.npmrc` token (12h) + adds `@mesh-tech` scope to project `.npmrc` + **strips an unscoped CodeArtifact `registry=` line from `~/.npmrc`** (after writing `~/.npmrc.bak-<ts>`) — that line is what a raw CodeArtifact sign-in *without* `--namespace @mesh-tech` leaves behind, and it makes CodeArtifact your DEFAULT registry so every *public* install 401s at the 12h expiry. Skipped under `--ci`, where the unscoped login is deliberate. `status` probes the registry rather than grepping the file: expired/missing token or a hijack exits 1, an unreachable registry warns and exits 0. **Step 0 is the Zitadel-gated broker:** with a context (`mesh registry login mesh.dev`) the CLI POSTs the cached id token to the platform's `registry-auth` service, which returns a CodeArtifact read token minted with its OWN AWS identity — so this path needs no AWS account, CLI or profile at all. It writes the same two `~/.npmrc` lines the AWS path does. Falls back to the AWS chain when the platform publishes no broker, when the broker is unreachable, or with `--no-broker`; a 403 (a valid Mesh account without the `registry:read` role) is fatal and does **not** fall through. `--publish` and `--profile` deliberately skip the broker — both are AWS-identity paths. `--profile mesh-dev` auto-runs the AWS SSO login when the session is stale. Also runs automatically after `mesh login {tenant}.{env}` (F2). `login --publish` assumes the registry's publisher role (from `--role` or the registry SSM export); `publish` pushes package snapshots. | `login`: `--no-broker`, `--profile <sso>`, `--role <arn>`, `--ci`, `--skip-npmrc`, `--publish`; `publish`: `--snapshot` |
+| `mesh registry` | `login [context]`, `logout`, `status`, `publish` | CodeArtifact auth → refreshes `~/.npmrc` token (12h) + adds `@mesh-tech` scope to project `.npmrc` + **strips an unscoped CodeArtifact `registry=` line from `~/.npmrc`** (after writing `~/.npmrc.bak-<ts>`) — that line is what a raw CodeArtifact sign-in *without* `--namespace @mesh-tech` leaves behind, and it makes CodeArtifact your DEFAULT registry so every *public* install 401s at the 12h expiry. Skipped under `--ci`, where the unscoped login is deliberate. `status` probes the registry rather than grepping the file: expired/missing token or a hijack exits 1, an unreachable registry warns and exits 0. **Step 0 is the Zitadel-gated broker:** the registry is ONE global identity (`utils/registry-identity.ts`, seeded from the compiled-in first-party alias table, session cached under the reserved key `registry`), so no context is needed — `mesh registry login` signs in to the registry's issuer if nothing is cached (browser; `--device` when headless; reuses a `mesh.dev` session on the same issuer) and POSTs the id token to the platform's `registry-auth` service, which returns a CodeArtifact read token minted with its OWN AWS identity — so this path needs no AWS account, CLI or profile at all. It writes the same two `~/.npmrc` lines the AWS path does. Falls back to the AWS chain when the platform publishes no broker, when the broker is unreachable, or with `--no-broker`; a 403 (a valid Mesh account without the `registry:read` role) is fatal and does **not** fall through. `--publish` and `--profile` deliberately skip the broker — both are AWS-identity paths. `--profile mesh-dev` auto-runs the AWS SSO login when the session is stale. `mesh login` never touches the registry (it prints a one-line pointer to `mesh registry login` when the machine has no registry access); `logout` clears the registry session only. A `[context]` on the read path is ignored with a notice; it still selects the SSM export for `--publish`. `login --publish` assumes the registry's publisher role (from `--role` or the registry SSM export); `publish` pushes package snapshots. | `login`: `--device`, `--no-broker`, `--profile <sso>`, `--role <arn>`, `--ci`, `--skip-npmrc`, `--publish`; `publish`: `--snapshot` |
 | `mesh db` | `connect`, `credentials`(alias `creds`), `env`, `psql`, `exec <command...>` | Access tenant DBs through an SSM bastion tunnel. `env` prints `export`s for `eval`; `exec` runs a command with `DATABASE_URL` injected. | `-t/--tenant`, `-e/--env`, `-a/--app`, `--app-tenant`, `--app-stage`; `exec` also: `--secret <arn>`, `--ssl <require\|no-verify\|disable>`, `--port <n>` |
 | `mesh deploy ...` | passthrough | Runs `pulumi` with the stack's `mesh:deployerRole` assumed. Unknown opts/args pass through; defaults to `up` if no subcommand. | `-s/--stack <name>` (stripped before pulumi `up`); everything else → pulumi (`preview`, `up --yes`, `destroy`, `stack output …`) |
-| `mesh dev` | `logs <service>`, `restart <service>`, `list`, `test-user [name]` | Two modes. **Tethered** (deployed stack): reads the Pulumi stack dev output, opens tunnels, injects Secrets Manager secrets; needs VPN/AWS creds. **Local** (`--local`, auto when the app has no `Pulumi.yaml`): synthesizes services from the app layout (subdirs with a `dev` script) and wires them to the `mesh start` platform (Temporal `local-dev`, app DB, Zitadel, ministack via `AWS_ENDPOINT_URL`) — zero AWS/VPN/Pulumi. `--externals` (local mode; `--mock` is an alias) also realizes the app's declared external services (package.json → `mesh.externals`; legacy `mesh.mocks` still read). Each declaration runs in one of three MODES — `mock` (the service is emulated: `openapi` → a Prism mock server from a spec, or `src` → a mock process), `local` (a local version of the real service: a `compose` file + published `port`, e.g. a vendor-shaped DB — up `--wait` before the app's services; a realization another checkout already started is **adopted** instead of failing — declared port already served **by a container this compose file defines** → adopt as-is (a publisher the file doesn't define — a native install or unrelated container — is a hard **conflict**, never adopted, since credentials would seed against the wrong service); a stopped foreign container holding the compose `container_name` → `docker start` + adopt — so startup order across checkouts doesn't matter; adopted externals stay out of session state and `--kill` tears down only externals this session created, volumes kept), `remote` (connect to the ACTUAL service: inline vendor `credentials`, or the credentials the app's tenant environment is configured with via `remote.env`). A declaration may carry SEVERAL realizations; `mode` picks the default and `--externals name=mode` overrides it per run (`--externals plaid-db=remote,plaid`). Remote externals get a blackbox uptime probe against their real endpoint. Failing to realize a remote external is fatal only when you named it; a bare `--externals` warns and keeps going, so the local loop still works with no AWS identity. Every realization seeds its ExternalService credential secret (`{{url}}`/`{{host}}`/`{{port}}` from the stand-in's endpoint, `{{env:VAR}}` from the shell) so `resolveCredentials()` runs the deployed code path against the stand-in. Runners: tmux (default) or `--runner docker` (CI/headless — services run as a docker compose project with host networking; parity `--status`/`--kill`/`logs`/`restart`). | root: `--local`, `--externals [name[=mode],…]` (alias `--mock`), `--runner <tmux\|docker>`, `--app <path>`, `--stage`, `--headless`, `--kill`, `--status`, `--json`, `--session <name>`, `--profile <sso>`; `logs`: `--tail <n>`; `test-user`: `--tenant/--env/--region` |
-| `mesh start` / `mesh stop` / `mesh status` | — | Full-local Mesh platform via docker compose (project `mesh-local`, assets ship in the CLI package): Postgres, Temporal (+UI), Zitadel, SpiceDB, ministack (local AWS fabric: SSM/Secrets/S3 registry), memcached, OpenSearch, a **local mailbox** (Mailpit — SMTP :1025 wired into Zitadel, inbox at http://localhost:8025, so user-activation and password-reset mail is testable locally instead of vanishing), **and the hosted observability backends** (OTel collector + Loki + Tempo + Prometheus — the Hub's logs/traces/metrics views run unchanged). Seeds tenant `local`/env `dev` on first boot and reconciles every Zitadel-provisioned app tenant into the registry. **No AWS creds, VPN, or Pulumi state needed — Docker only.** The **Hub is included by default** — three overlay services, `hub-api` + `hub-ui` + `hub-auth` (an oauth2-proxy; it is what binds the front-door port and proxies to `hub-ui`, so front-door 401s/redirect loops are in ITS logs, not `hub-api`'s) — the front door at http://localhost:9000 — `MESH_HUB_PORT=<port>` publishes it elsewhere when 9000 is taken; export it for `mesh status` too (images build once from the published @mesh-tech/hub tarball; needs `mesh registry login` the first time; `--no-hub` opts out). Auth contract (MESH-2479): `--with-hub` refreshes to the latest published hub after preflighting the CodeArtifact token — with an expired/missing token it falls back to the already-built local images (warn + fixing command), or fails fast naming `mesh registry login --profile mesh-dev` when none exist; the default (no-flag) path with an expired/missing token starts hub-less and names the same fix. `--hub-from-source` sidesteps the registry entirely — it builds the Hub images from THIS checkout's `apps/hub`, which is the only way to run an unpublished Hub change in the containerized stack. Non-TTY runs (CI, agents, piped output) get an `[HH:MM:SS]` prefix on every step line plus a heartbeat during long fetches/builds — `MESH_LOG_TIMESTAMPS=1/0` forces the prefix on/off (default: on when stderr is not a TTY). | `start`: `--no-seed`, `--no-hub`, `--with-hub`, `--hub-from-source`; `stop`: `--destroy` (drops volumes/seeds); `status`: `--json` |
+| `mesh dev` | `logs <service>`, `restart <service>`, `list`, `test-user [name]` | Two modes. **Tethered** (deployed stack): reads the Pulumi stack dev output, opens tunnels, injects Secrets Manager secrets; needs VPN/AWS creds. **Local** (`--local`, auto when the app has no `Pulumi.yaml`): synthesizes services from the app layout (subdirs with a `dev` script, plus a synthesized `dg dev` webserver when the app root is a Dagster workspace — `dg.toml`, or `[tool.dg]` in `pyproject.toml`; such an app runs with zero package.json services — see `docs/guides/local-dagster.md`) and wires them to the `mesh start` platform (Temporal `local-dev`, app DB, Zitadel, ministack via `AWS_ENDPOINT_URL`) — zero AWS/VPN/Pulumi. `--externals` (local mode; `--mock` is an alias) also realizes the app's declared external services (package.json → `mesh.externals`; legacy `mesh.mocks` still read). Each declaration runs in one of three MODES — `mock` (the service is emulated: `openapi` → a Prism mock server from a spec, or `src` → a mock process), `local` (a local version of the real service: a `compose` file + published `port`, e.g. a vendor-shaped DB — up `--wait` before the app's services; a realization another checkout already started is **adopted** instead of failing — declared port already served **by a container this compose file defines** → adopt as-is (a publisher the file doesn't define — a native install or unrelated container — is a hard **conflict**, never adopted, since credentials would seed against the wrong service); a stopped foreign container holding the compose `container_name` → `docker start` + adopt — so startup order across checkouts doesn't matter; adopted externals stay out of session state and `--kill` tears down only externals this session created, volumes kept), `remote` (connect to the ACTUAL service: inline vendor `credentials`, or the credentials the app's tenant environment is configured with via `remote.env`). A declaration may carry SEVERAL realizations; `mode` picks the default and `--externals name=mode` overrides it per run (`--externals plaid-db=remote,plaid`). Remote externals get a blackbox uptime probe against their real endpoint. Failing to realize a remote external is fatal only when you named it; a bare `--externals` warns and keeps going, so the local loop still works with no AWS identity. Every realization seeds its ExternalService credential secret (`{{url}}`/`{{host}}`/`{{port}}` from the stand-in's endpoint, `{{env:VAR}}` from the shell) so `resolveCredentials()` runs the deployed code path against the stand-in. Runners: tmux (default) or `--runner docker` (CI/headless — services run as a docker compose project with host networking; parity `--status`/`--kill`/`logs`/`restart`). | root: `--local`, `--externals [name[=mode],…]` (alias `--mock`), `--runner <tmux\|docker>`, `--app <path>`, `--stage`, `--headless`, `--kill`, `--status`, `--json`, `--session <name>`, `--profile <sso>`; `logs`: `--tail <n>`; `test-user`: `--tenant/--env/--region` |
+| `mesh start` / `mesh stop` / `mesh status` | — | Full-local Mesh platform via docker compose (project `mesh-local`, assets ship in the CLI package): Postgres, Temporal (+UI), Zitadel, SpiceDB, ministack (local AWS fabric: SSM/Secrets/S3 registry; also seeds a general-purpose `mesh-local-data` bucket with sample data — see `docs/guides/local-dagster.md`), memcached, OpenSearch, a **local mailbox** (Mailpit — SMTP :1025 wired into Zitadel, inbox at http://localhost:8025, so user-activation and password-reset mail is testable locally instead of vanishing), **and the hosted observability backends** (OTel collector + Loki + Tempo + Prometheus — the Hub's logs/traces/metrics views run unchanged). Seeds tenant `local`/env `dev` on first boot and reconciles every Zitadel-provisioned app tenant into the registry. **No AWS creds, VPN, or Pulumi state needed — Docker only.** The **Hub is included by default** — three overlay services, `hub-api` + `hub-ui` + `hub-auth` (an oauth2-proxy; it is what binds the front-door port and proxies to `hub-ui`, so front-door 401s/redirect loops are in ITS logs, not `hub-api`'s) — the front door at http://localhost:9000 — `MESH_HUB_PORT=<port>` publishes it elsewhere when 9000 is taken; export it for `mesh status` too (images build once from the published @mesh-tech/hub tarball; needs `mesh registry login` the first time; `--no-hub` opts out). Auth contract (MESH-2479): `--with-hub` refreshes to the latest published hub after preflighting the CodeArtifact token — with an expired/missing token it falls back to the already-built local images (warn + fixing command), or fails fast naming `mesh registry login --profile mesh-dev` when none exist; the default (no-flag) path with an expired/missing token starts hub-less and names the same fix. `--hub-from-source` sidesteps the registry entirely — it builds the Hub images from THIS checkout's `apps/hub`, which is the only way to run an unpublished Hub change in the containerized stack. Non-TTY runs (CI, agents, piped output) get an `[HH:MM:SS]` prefix on every step line plus a heartbeat during long fetches/builds — `MESH_LOG_TIMESTAMPS=1/0` forces the prefix on/off (default: on when stderr is not a TTY). | `start`: `--no-seed`, `--no-hub`, `--with-hub`, `--hub-from-source`; `stop`: `--destroy` (drops volumes/seeds); `status`: `--json` |
 | `mesh hub` | `dev` | One-command local Hub over a running dev-local stack. Launches the CURRENT-CODE Hub (apps/hub api+ui from a mesh-platform checkout) pointed at an existing `mesh dev` session — auto-assembles `HUB_TENANT`/`HUB_SCOPE_ENV`/`HUB_SCOPE_TENANTS` (session platform context), `TEMPORAL_ADDRESS` (session tunnel), `AWS_*` (session scratch profile), `DEV_USER_*`+`DEV_USER_TOKEN_URL` (session dev identity; starts its own token-server if the session predates it, so tokens stay fresh past ~1h). Refuses to launch with an empty tenant scope (which the Hub renders as silently blank). Runs in tmux session `mesh-hub-dev`. | `dev`: `--session <name>`, `--tenants <csv>`, `--port` (or `MESH_HUB_DEV_PORT`, default 9000; the containerized `mesh start` Hub uses the separate `MESH_HUB_PORT`), `--api-port <3002>`, `--platform-dir <dir>` (or `MESH_PLATFORM_DIR`), `--print-env`, `--kill` |
 | `mesh create-app` | — | Scaffold a tenant app. Composable mode (primitives) or legacy template mode. Interactive when TTY + missing flags. Run at the root of an **empty** `{tenant}-mesh-apps` git clone (a `.git` with no `package.json`/`apps/`/`tenants/` and no enclosing pnpm workspace), it first generates the repo-level shape — `pnpm-workspace.yaml` over `apps/*`, private root `package.json`, scoped-registry `.npmrc`, base `tsconfig.json`, `.gitignore`, `README.md`, `apps/` — then scaffolds the app. Never overwrites an existing file. | `--tenant`, `--name`, `--primitives <csv>` (`service,database,temporal,bucket`), `--template <workflow\|api-auth\|api-role-gating\|external-service>`, `--test` (writes to `tests/tenants/`) |
 | `mesh app` | `check [apps...]` | Checks tenant apps against the **Mesh app contract** (`@mesh-tech/app-kit#apps` → `references/app-contract.md`): a UI is its own Service (`UI_IN_API`, `NO_UI_SERVICE`), the app registers with the Hub (`NO_REGISTER`), sign-in goes through the platform proxy (`CUSTOM_SESSION_AUTH`), logs go through `@mesh-tech/logger` (`NON_PLATFORM_LOGGER`), and people/roles/keys are surfaced through the Hub (`NO_AUTHZ_POINTER`, `METADATA_UNPUBLISHED`, `IAC_GRANTS`, `PASSWORD_STORE`, `EMAIL_ALLOWLIST`, `USERS_TABLE`); a tenant-local auth lib is advisory (`LOCAL_AUTH_LIB`). One check per code on the `mesh dev doctor` engine; exit 1 on any BLOCK. The SAME gate `mesh create-app` runs after scaffolding, tenant CI runs beside `mesh skills sync --check`, and the platform reviewer runs on a PR. Default: every `apps/*` with a Pulumi program. | `[apps...]` (repo-relative dirs), `--root <dir>`, `--json`, `--verbose` |
-| `mesh init` | `app-tenant` | The repo doctor (A1–A3): read-only check pipeline — CLI auth, registry access, public npm not hijacked, platform reachable (Hub API), app-tenant registration, deployer role, repo shape, agent skills — each failure names the exact `mesh` fix. **Registry access** is a live probe of the CodeArtifact token, not a grep of `~/.npmrc`: an expired token `fail`s (it used to `pass`), an unreachable registry `warn`s (a network fault is not an auth fault). **Public npm not hijacked** fails when `~/.npmrc` carries an unscoped `registry=<codeartifact>` line, which makes CodeArtifact your default registry — one `mesh registry login` fixes both. Exit 0 iff all pass; re-run any time. `--fix` applies developer-scope fixes (registry token, skills sync); operator actions stay remediations — and `--profile <p>` is the SSO profile `--fix` logs into CodeArtifact with, echoed back in the `fix:` command each check prints so the advertised command is the one that runs. | `app-tenant`: `--tenant <name>`, `--context <ctx>` (default `local`), `--hub-url <url>`, `--fix`, `--profile <sso>`, `--json` |
+| `mesh init` | (bare = wizard), `app-tenant`, `platform <tenant>` | **Bare `mesh init` is the guided setup wizard**: tenant (pre-filled from `mesh.json` / Pulumi `mesh:tenant`), local-only vs deployed platform, registry access (in-process `mesh registry login`; a 403 marks the step ✘ and the wizard continues), platform sign-in on the deployed branch only, repo bootstrap (empty clone) or shape check, skills sync; writes `mesh.json` (`{tenant, platform}`) so `create-app` / `mesh dev` default `--tenant`; flags `--tenant --local\|--platform <env> --device --profile --skip-repo --yes --json`; exit 1 if any step failed, 130 on Ctrl-C. `app-tenant` is the repo doctor (A1–A3): read-only check pipeline — CLI auth, registry access, public npm not hijacked, platform reachable (Hub API), app-tenant registration, deployer role, repo shape, agent skills — each failure names the exact `mesh` fix. **Registry access** is a live probe of the CodeArtifact token, not a grep of `~/.npmrc`: an expired token `fail`s (it used to `pass`), an unreachable registry `warn`s (a network fault is not an auth fault). **Public npm not hijacked** fails when `~/.npmrc` carries an unscoped `registry=<codeartifact>` line, which makes CodeArtifact your default registry — one `mesh registry login` fixes both. Exit 0 iff all pass; re-run any time. `--fix` applies developer-scope fixes (registry token, skills sync); operator actions stay remediations — and `--profile <p>` is the SSO profile `--fix` logs into CodeArtifact with, echoed back in the `fix:` command each check prints so the advertised command is the one that runs. | `app-tenant`: `--tenant <name>`, `--context <ctx>` (default `local`), `--hub-url <url>`, `--fix`, `--profile <sso>`, `--json` |
 | `mesh skills` | `sync` | Agent-skill distribution (D1/D2): installs the base building-with-Mesh skills into the repo's `.claude/skills/mesh-*/` (managed-marker files) and wires TanStack-Intent discovery for the pattern skills shipping inside `@mesh-tech/*` packages (`.intent/hooks/` gate + `.claude/settings.json` SessionStart hook + AGENTS.md fence — the PR #2356 mechanism, vendored). Idempotent; runs automatically after `mesh create-app`. | `sync`: `--check` (CI/doctor: exit 1 when missing/stale), `--root <path>` |
 | `mesh docs` | `build`, `portal`, `cli-reference` | `build` builds/validates an **app's** OpenAPI specs from `docs/docs.config.json` (file copy or SymXchange generation). `start`/`stop`/`list` are the simple front door: `start` serves the docs on loopback — DETACHED in a tmux session named `mesh-docs` by default (prints the bare URL once live; `stop` kills the session; `--foreground` or any non-TTY runs in the foreground) — serving the working tree in a mesh-platform checkout, or the published `@mesh-tech/docs` artifact fetched from the role-gated CodeArtifact registry anywhere else (docs version == the @mesh-tech/* baseline it describes; `-v` pins, default latest; cached under `~/.cache/mesh/docs/`). `list` shows the published docs versions. `portal` assembles the **Mesh developer portal** (docs.meshtech.io) from every `docs.json`-opted-in doc root in the repo — a directory publishes its markdown iff it holds a `docs.json`; route/title/order/nav derive from the tree and each file, so adding a doc is adding one file — then hands the assembled tree to Zudoku (`apps/docs/`): default runs `zudoku build`, `--serve` runs `zudoku dev`, `--assemble-only`/`-o` stop after assembly. Links between published files are rewritten to routes; links to unpublished repo files become blob links. Reserved directory names (`plans/`, `designs/`, `incidents/`, … the frozen list in `src/docs/schema.ts`) are excluded at any depth. `cli-reference` regenerates `docs/portal/generated/cli-reference.md` from THIS command tree, so it can't document a flag the binary lacks. `portal --check` is the CI gate (`pnpm check:docs-portal`): docs.json schema validation, reserved-segment fail-close, broken links, and CLI-reference staleness; `--diff-base <ref>` prints the publish-set diff for PR job summaries. Hidden commands and `--version` are excluded from the reference. | `build`: `-c/--config <path>` (default `docs/docs.config.json`); `portal`: `-o/--out <dir>`, `--assemble-only`, `--serve`, `-p/--port <n>` (default `3000`), `--check`, `--print-manifest`, `--diff-base <ref>`, `--manifest-out <path>`; `cli-reference`: `-o/--out <path>`, `--check`; `start`: `-v/--version <v>` (default: latest), `-p/--port <n>` (default `4400`; `0` is foreground-only), `--dev` (foreground HMR), `--foreground`; `stop`: no flags; `list`: no flags |
 | `mesh stack` | `init`, `rm <name>` | `init` creates a personal dev stack `dev-{github-user}`, copying base-stack config, setting `mesh:deploy: false`, inheriting KMS secrets provider; `--worktree` appends the git-worktree token (`dev-{user}-{token}`) so concurrent worktrees deploy to **distinct** stacks (SSM paths, Temporal namespace, Nexus endpoint all derive from the stack name → one discriminator isolates all three; no-op on the primary checkout). `rm <name>` removes a personal stack (backend state + local config), credentialed. Both assume the deployer role — no raw `pulumi` / manual creds needed. | `init`: `--from <stack>`, `--name <stack>`, `--worktree`; `rm`: `--yes` |
@@ -119,10 +119,11 @@ Top-level commands (each `register*` is wired in `src/program.ts`, which `bin/me
 ### 1. Onboard: auth + install dependencies
 
 ```bash
-mesh login mesh.dev          # OIDC; auto-uses device flow in remote/headless envs
-mesh login mesh.dev --device # FORCE device flow (browser callback times out over SSH/tailscale)
-mesh registry login mesh.dev # registry token via the Zitadel broker (no AWS needed) + project .npmrc scope
+mesh registry login          # ONE command: signs in (browser) + registry token via the Zitadel broker + project .npmrc scope — no AWS, no platform context
+mesh registry login --device # same, device-code sign-in (browser callback times out over SSH/tailscale)
+mesh init                    # guided: tenant, local-only vs deployed, registry access, repo bootstrap/doctor, skills → mesh.json
 pnpm install                 # now resolves @mesh-tech/* from CodeArtifact
+# mesh login <tenant>.<env> is for a DEPLOYED platform only (mesh deploy / Hub / VPN) — never needed for packages
 ```
 
 ### 1b. Run the platform fully locally (no cloud)

package/templates/apps-repo/gitignore

@@ -37,5 +37,9 @@ pnpm-debug.log*
 # Personal dev stacks (never commit) — keep the shared dev/prod stacks
 Pulumi.dev-*.yaml
 
+# Dagster local instance state (`mesh dev --local` pins DAGSTER_HOME here)
+.dagster/
+.tmp_dagster_home_*/
+
 # Temporary
 temp/