@toolbeltai/skills 0.2.4 → 1.0.6

package/README.md

@@ -1,63 +1,58 @@
 # @toolbeltai/skills
 
-> Official Toolbelt skills for Claude Code and other MCP-capable agents —
-> SQL, vectors, graphs, geospatial, streaming, all as named slash commands.
+The official skill for [Toolbelt](https://toolbelt.ai). Installs one
+`/toolbelt` skill that any MCP-capable agent can use to connect to a
+Toolbelt namespace and hand off to Toolbelt's MCP tools.
 
-## Install (standalone)
+[![npm version](https://img.shields.io/npm/v/@toolbeltai/skills.svg)](https://www.npmjs.com/package/@toolbeltai/skills)
+[![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
+
+## Install
 
 ```bash
 npx @toolbeltai/skills install
 ```
 
-Copies every skill to `~/.claude/skills/` (flat, per the AgentSkills spec).
-Restart Claude Code and you'll see them as slash commands: `/toolbelt-start`,
-`/toolbelt-geo`, …
-
-No account required; no network calls to Toolbelt. Skills work against any
-Toolbelt MCP server (cloud or self-hosted).
+Copies `toolbelt/SKILL.md` into `~/.claude/skills/toolbelt/` (flat
+layout per the [AgentSkills spec](https://agentskills.io)). Restart your
+client; the skill appears as `/toolbelt`.
 
-## Install (via Toolbelt CLI)
+No Toolbelt account required — the skill provisions one on first use.
 
-If you want the skills **and** a preconfigured MCP connection in one step:
+## What the skill does
 
-```bash
-npx @toolbeltai/cli
-```
+When an agent runs `/toolbelt` for the first time, it:
 
-Provisions an anonymous Toolbelt account, registers the MCP server in your
-agent, and installs these skills — all at once.
+1. Checks for an existing Toolbelt MCP connection.
+2. If none, calls the public `/api/onboard` endpoint to provision a
+   free anonymous account.
+3. Writes the MCP connection into the agent's client config.
+4. Hands off to Toolbelt's MCP tools for the actual work.
 
-## Skills
-
-| Skill | Command | What it does |
-| --- | --- | --- |
-| [toolbelt-start](toolbelt-start/) | `/toolbelt-start` | Onboard end-to-end — provision, ingest, first query |
-| [toolbelt-analyze](toolbelt-analyze/) | `/toolbelt-analyze` | Upload 1+ CSVs, ask in plain English, get SQL answers (single-table or multi-table JOIN) |
-| [toolbelt-find](toolbelt-find/) | `/toolbelt-find` | Upload a document, retrieve passages by semantic similarity |
-| [toolbelt-entities](toolbelt-entities/) | `/toolbelt-entities` | Auto-extract entities and relationships from docs, explore with Cypher |
-| [toolbelt-geo](toolbelt-geo/) | `/toolbelt-geo` | GPU-accelerated geospatial — distance, containment, routing, map rendering |
-| [toolbelt-stream](toolbelt-stream/) | `/toolbelt-stream` | Connect Kafka, aggregate over windows, detect anomalies |
-| [toolbelt-invite](toolbelt-invite/) | `/toolbelt-invite` | Emit a connection URL so another agent can join this workspace |
+Full playbook in [toolbelt/SKILL.md](toolbelt/SKILL.md).
 
 ## Works with
 
-Claude Code · OpenClaw · Cursor · Gemini CLI · Codex CLI · Windsurf · any MCP client
+Claude Code · Claude Desktop · OpenClaw · Cursor · Gemini CLI · Codex
+CLI · Windsurf · any MCP client.
 
 ## Commands
 
 ```bash
-npx @toolbeltai/skills install       # install all skills
-npx @toolbeltai/skills uninstall     # remove them
-npx @toolbeltai/skills list          # list what would be installed
+npx @toolbeltai/skills install       # install
+npx @toolbeltai/skills uninstall     # remove
+npx @toolbeltai/skills list          # show what would install
 npx @toolbeltai/skills path          # print install target
 ```
 
-## Docs
+## Links
 
-- Skill reference: <https://docs.toolbelt.ai>
-- Toolbelt CLI: <https://github.com/toolbeltai/toolbelt>
-- Releasing: [RELEASING.md](./RELEASING.md)
+- Site: <https://toolbelt.ai>
+- Docs: <https://toolbelt.ai/docs>
+- App / billing: <https://app.toolbelt.ai>
 - Contributing: [CONTRIBUTING.md](./CONTRIBUTING.md)
+- Releasing: [RELEASING.md](./RELEASING.md)
+- Support: <support@toolbelt.ai>
 
 ## License
 

package/bin/install.js

@@ -26,11 +26,16 @@ const PKG_ROOT = join(dirname(fileURLToPath(import.meta.url)), '..');
 const TARGET = join(homedir(), '.claude', 'skills');
 const LEGACY_NESTED_DIR = join(TARGET, 'toolbelt');
 
-// Pre-0.2.0 skill folder names. Removed on install so upgraders don't
-// see ghost slash commands for skills that were renamed or merged.
-// Added 2026-04-23 with the toolbelt-* prefix rename; safe to drop this
-// list in a future major once the userbase has cycled.
+// Skill folder names that earlier versions of this package installed
+// but no longer ship. Removed on install so upgraders don't see ghost
+// slash commands for skills that were renamed, merged, or collapsed.
+//
+//   v0.1.x → v0.2.0   prefix rename + a couple of merges
+//   v0.2.x → v1.0.0   seven task skills collapsed into one `toolbelt`
+//
+// Safe to drop these in a future major once the userbase has cycled.
 const LEGACY_SKILL_NAMES = [
+  // pre-v0.2.0
   'run-toolbelt',
   'sql-analyst',
   'data-blend',
@@ -39,6 +44,14 @@ const LEGACY_SKILL_NAMES = [
   'geo-analyst',
   'streaming-analyst',
   'multi-agent-workspace',
+  // pre-v1.0.0 (the seven that collapsed into the flagship)
+  'toolbelt-start',
+  'toolbelt-analyze',
+  'toolbelt-find',
+  'toolbelt-entities',
+  'toolbelt-geo',
+  'toolbelt-stream',
+  'toolbelt-invite',
 ];
 
 /** Any top-level directory containing a SKILL.md is a skill. */

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@toolbeltai/skills",
-  "version": "0.2.4",
-  "description": "Official Toolbelt skills — named /toolbelt-* flows (start, analyze, find, entities, geo, stream, invite) that teach any MCP-capable agent to orchestrate Toolbelt's MCP tools end-to-end.",
+  "version": "1.0.6",
+  "description": "Official Toolbelt skill — one /toolbelt slash command that gives any MCP-capable agent a shared brain for your data. Detects or provisions a free Toolbelt account, configures the MCP connection, and hands off to Toolbelt's MCP tools (vector, knowledge graph, SQL, geospatial, streaming, timeline).",
   "license": "MIT",
   "homepage": "https://toolbelt.ai",
   "repository": {
@@ -31,14 +31,7 @@
   },
   "files": [
     "bin",
-    "assets",
-    "toolbelt-start",
-    "toolbelt-analyze",
-    "toolbelt-find",
-    "toolbelt-entities",
-    "toolbelt-geo",
-    "toolbelt-stream",
-    "toolbelt-invite",
+    "toolbelt",
     "LICENSE",
     "README.md"
   ],

package/toolbelt/SKILL.md

@@ -0,0 +1,288 @@
+---
+name: toolbelt
+description: >
+  Toolbelt is a collaborative substrate for your agents — one shared brain
+  for your data. Connect AI agents to vector search, knowledge graph, SQL,
+  geospatial, and streaming data through one MCP server. Findings,
+  decisions, and observations persist across sessions and across agents.
+
+  Use this skill at the start of any task where an agent needs persistent
+  memory across sessions, natural-language access to structured or
+  unstructured data, or shared state with other agents working on the same
+  data. The skill handles first-time setup: provisions a free Toolbelt
+  account if none exists, configures the MCP connection in the agent's
+  client, optionally upgrades the account by email, and hands off to
+  Toolbelt's MCP tools for the actual work.
+
+  NOT for one-off lookups that don't benefit from persistent memory or
+  structured data — use the agent's native tools for those.
+license: MIT
+compatibility: >
+  Requires an MCP-compatible AI agent (Claude Code, Claude Desktop,
+  OpenClaw, Cursor, Windsurf, Gemini CLI, Codex CLI, or any client that
+  supports MCP server connections). No Toolbelt account required — this
+  skill provisions one on first use.
+version: "1.0.0"
+metadata:
+  author: toolbeltai
+  homepage: "https://toolbelt.ai"
+---
+
+# Toolbelt — A collaborative substrate for your agents
+
+**Your data. Your agents. One shared brain.**
+
+Toolbelt connects AI agents to vector search, knowledge graph, SQL,
+geospatial, and streaming data through one MCP server. Findings,
+decisions, and observations persist across sessions and across agents.
+
+## Two surfaces — keep them straight
+
+Toolbelt has exactly two surfaces. Knowing which is which is the most
+important thing in this skill:
+
+| Surface | URL | Who uses it | When |
+|---|---|---|---|
+| **MCP server** (the agentic surface) | `https://mcp.toolbelt.ai/mcp` | **Agents** | Every data operation. This is where the agentic flow happens — search, SQL, knowledge graph, record findings, read timeline. |
+| **app.toolbelt.ai** (the human web UI) | `https://app.toolbelt.ai` | **Humans** (in a browser) | Sign in, view/manage namespaces, billing, Pro/Team upgrade. Plus a small HTTP API at `/api/onboard*` used **once** during setup. |
+
+**Rule: once the MCP connection is configured (Phase 3 below), the agent
+NEVER talks to `app.toolbelt.ai` again** — every subsequent action goes
+through MCP. The only reason to mention `app.toolbelt.ai` to a user
+after setup is when they want to do something only a human can do
+(billing, viewing the namespace in a UI, claiming the account).
+
+## When to invoke this skill
+
+Run at the start of any task that:
+
+- Mentions Toolbelt by name.
+- Needs persistent memory across turns or sessions.
+- Needs natural-language access to structured or unstructured data.
+- Involves multiple agents collaborating on the same data.
+- Would otherwise require wiring up several separate MCP tools.
+
+## Phases — execute in order, no human prompts unless explicitly stated
+
+### Phase 1 — Detect existing connection
+
+Try calling the Toolbelt MCP tool `toolbelt_list_namespaces`.
+
+- Returns successfully → user is already connected → skip to **Phase 4**.
+- Tool unavailable or returns auth error → continue to **Phase 2**.
+
+### Phase 2 — Provision a free Toolbelt account (one HTTP call)
+
+This is the only time the agent hits `app.toolbelt.ai`'s HTTP API
+directly. Anonymous; no auth required.
+
+```http
+POST https://app.toolbelt.ai/api/onboard
+Content-Type: application/json
+
+{}
+```
+
+Response shape (verify against this exact structure):
+
+```json
+{
+  "success": true,
+  "user": { "id": "@anon_..." },
+  "namespace": { "id": "<uuid>", "name": "My Namespace" },
+  "mcpUrl": "https://mcp.toolbelt.ai/mcp",
+  "token": "tb_...",
+  "expiresAt": "<ISO timestamp>"
+}
+```
+
+Capture: `token`, `mcpUrl`, `user.id`, `namespace.id`, `expiresAt`. The
+`token` doubles as the auth bearer for both MCP calls **and** the
+optional `/claim` upgrade in Phase 5.
+
+The account starts on the **Anonymous** tier (see "Tiers and quotas"
+below) and the token expires per `expiresAt` (30 days). Claim by email
+in Phase 5 to make it persistent.
+
+### Phase 3 — Configure the agent's MCP client (one-time)
+
+Write the MCP connection into the agent's client config. The shape is
+identical across clients — only the file path differs.
+
+**Claude Code (CLI):**
+```bash
+claude mcp add toolbelt \
+  --transport http <mcpUrl> \
+  --header "Authorization: Bearer <token>"
+```
+
+**Claude Desktop / Cursor / Windsurf / OpenClaw / Gemini CLI / Codex CLI** —
+add to the client's MCP config JSON:
+
+```json
+{
+  "mcpServers": {
+    "toolbelt": {
+      "url": "<mcpUrl>",
+      "headers": { "Authorization": "Bearer <token>" }
+    }
+  }
+}
+```
+
+Config file locations:
+
+| Client | Path |
+|---|---|
+| Claude Desktop (macOS) | `~/Library/Application Support/Claude/claude_desktop_config.json` |
+| Claude Desktop (Windows) | `%APPDATA%\Claude\claude_desktop_config.json` |
+| Cursor | `~/.cursor/mcp.json` |
+| Windsurf | `~/.codeium/windsurf/mcp_config.json` |
+| OpenClaw | `~/.openclaw/mcp.json` |
+| Gemini CLI | `~/.gemini/mcp.json` |
+| Codex CLI | `~/.codex/mcp.json` |
+
+Tell the user **once**: "Toolbelt is provisioned. Reload your MCP
+connection to activate it." Most clients pick up changes on next
+request; some need a restart.
+
+### Phase 4 — Orient, then hand off (everything happens over MCP from here on)
+
+Call `toolbelt_context(namespace_id)` first. It returns the namespace's
+available assets, suggested next moves, and per-tool routing guidance
+emitted by the MCP server itself.
+
+Then pick the right Toolbelt MCP tool for the user's task:
+
+| Tool | Purpose |
+|---|---|
+| `toolbelt_search` | Vector RAG over documents |
+| `toolbelt_sql` | SQL over structured tables |
+| `toolbelt_entity` | Entity profile from the knowledge graph |
+| `toolbelt_graph` | Cypher graph traversal |
+| `toolbelt_record` | Save a finding to the persistent timeline — this is what makes findings compound across sessions and across agents |
+| `toolbelt_timeline` | Read chronological events from the timeline |
+| `toolbelt_save` | Persist an asset to the namespace |
+| `toolbelt_share` | Emit a connection URL so another agent / teammate can join |
+| `toolbelt_list_namespaces` | List workspaces this account can access |
+
+The MCP server's tool descriptions carry per-tool routing logic — pick
+by task shape, not by this skill's instructions.
+
+### Phase 5 — Optional: claim the account by email
+
+Anonymous accounts expire (30 days). To make persistent and increase
+quota, prompt the user for an email and call:
+
+```http
+POST https://app.toolbelt.ai/api/onboard/claim
+Authorization: Bearer <token>
+Content-Type: application/json
+
+{"email": "user@example.com"}
+```
+
+User receives a verification email. Then:
+
+```http
+POST https://app.toolbelt.ai/api/onboard/claim/verify
+Authorization: Bearer <token>
+Content-Type: application/json
+
+{"code": "<code from email>"}
+```
+
+After verification the account is upgraded from **Anonymous** to
+**Verified** — same token, higher quota, persistent across sessions.
+
+## Tiers and quotas
+
+Match `toolbelt.ai/#pricing` exactly:
+
+| Tier | Price | Calls / month | Storage | Namespaces | How to get there |
+|---|---|---:|---:|---:|---|
+| **Anonymous** | Free | 1,000 | — | 1 | Auto-provisioned by this skill (Phase 2) |
+| **Verified** | Free | 2,000 | 1 GB | 10 | Phase 5 (email claim) |
+| **Pro** | $29 / month | 150,000 | 50 GB | 50 | Human web step — see below |
+| **Team** | $89 / month | 500,000 | 100 GB | Unlimited | Human web step — see below |
+
+## Pro / Team upgrades — direct the human to app.toolbelt.ai
+
+Stripe checkout requires a real browser session. **Agents cannot do
+this; do not pretend to.** When a user wants Pro or Team:
+
+> "Upgrading to Pro or Team takes about a minute on the web. Open
+> <https://app.toolbelt.ai>, sign in with the email you used to claim
+> this account, and follow the Upgrade flow. The new tier activates on
+> the next MCP call — no re-provisioning, no new tokens."
+
+Do not invent upgrade URLs. Do not collect credit card info. Do not
+prompt for billing data. The skill's job ends at "direct the human to
+the right page."
+
+## Output after Phase 4 succeeds
+
+Emit a brief connection status to the user:
+
+```yaml
+toolbelt_connection:
+  status: connected
+  mcp_url: <mcpUrl>
+  user_id: <user.id>
+  namespace_id: <namespace.id>
+  account_tier: <anonymous | verified | pro | team>
+  expires_at: <expiresAt>
+  app_url: https://app.toolbelt.ai
+```
+
+Then proceed with the user's actual task using the MCP tools.
+
+## Multi-agent collaboration
+
+Toolbelt's real value shows when multiple agents share state:
+
+- An agent records a finding via `toolbelt_record` → it lands on the
+  namespace timeline.
+- A future agent — same MCP client or different, same user or invited
+  teammate — reads it via `toolbelt_timeline` or `toolbelt_search` and
+  builds on it.
+- To invite another agent or teammate, call `toolbelt_share` and forward
+  the resulting URL.
+
+Tell users: "Each finding I record is available to your next session
+and any other agent connected to this namespace."
+
+## Reference URLs
+
+| Purpose | URL |
+|---|---|
+| Marketing site + pricing | <https://toolbelt.ai> |
+| Docs (concepts, tools, self-hosting) | <https://toolbelt.ai/docs> |
+| Human web UI (sign in, billing, namespace UI) | <https://app.toolbelt.ai> |
+| MCP endpoint (set in Phase 3) | `https://mcp.toolbelt.ai/mcp` |
+| Onboard API base (Phase 2 + 5 only) | `https://app.toolbelt.ai/api/onboard` |
+| Support | <support@toolbelt.ai> |
+
+## Common failure modes
+
+| Symptom | Cause | Handling |
+|---|---|---|
+| `toolbelt_list_namespaces` returns 401 | Stored MCP token expired or was revoked | Go back to Phase 2, provision a fresh anonymous account. |
+| Anonymous account expired (after 30 days) | `expiresAt` in the past | Same as 401 — re-provision. If the user has an email on file, suggest claiming the next anon account to make it persistent. |
+| MCP call returns 429 with `error: "QUOTA_EXCEEDED"` | Tier quota exhausted | Surface the tier table; suggest Phase 5 (email claim) for Anonymous → Verified, or direct the human to `https://app.toolbelt.ai` for Pro/Team. |
+| Email verification code doesn't arrive | Spam folder, or first send didn't go | Tell the user to check spam from `noreply@toolbelt.ai`, or call `POST /api/onboard/claim` again to re-send. |
+| `mcp.toolbelt.ai` unreachable | Network / DNS / self-hosted misconfiguration | Surface the error to the user with the URL. Don't attempt fallback — there's no fallback endpoint. |
+
+## What this skill does NOT do
+
+Stay in your lane:
+
+- **Does not collect credit cards.** Stripe is a browser flow.
+- **Does not generate or store passwords.** Authentication is by token,
+  managed by the MCP client config.
+- **Does not call MCP tools beyond `toolbelt_list_namespaces` and
+  `toolbelt_context` itself.** Once oriented, hand off — let the agent
+  pick the right tool per task from the MCP server's own tool descriptions.
+- **Does not invent endpoints.** Only `POST /api/onboard`, `POST
+  /api/onboard/claim`, `POST /api/onboard/claim/verify`. Everything else
+  is MCP.

package/assets/demos/fake-curl.sh

@@ -1,3 +0,0 @@
-#!/bin/bash
-sleep 1
-echo '{"mcpUrl": "https://mcp.toolbelt.ai/t/tb_1234567890/8bf91bad-02df-4137-97ac-69a006bac193/mcp"}'

package/assets/demos/toolbelt-geo-demo.sh

@@ -1,85 +0,0 @@
-#!/usr/bin/env bash
-# Simulates the toolbelt-geo skill running inside Claude Code
-
-G='\033[0;32m'   # green
-B='\033[0;34m'   # blue
-Y='\033[1;33m'   # yellow
-C='\033[0;36m'   # cyan
-D='\033[2;37m'   # dim
-W='\033[1;37m'   # bold white
-NC='\033[0m'
-
-printf "\n"
-printf "${D}> /toolbelt-geo${NC}\n"
-sleep 0.8
-
-printf "\n"
-printf "${C}● Phase 0  Verifying Toolbelt MCP connection...${NC}\n"
-sleep 0.6
-printf "${G}  ✓ Connected${NC}  |  namespace: demo-workspace\n"
-sleep 0.3
-
-printf "\n"
-printf "${C}● Phase 1  Resolving namespace...${NC}\n"
-sleep 0.4
-printf "${G}  ✓ Using${NC}  demo-workspace  (a3f2c1d9-4b2e-4a7c-9f1d-8e3b0c2d6a5f)\n"
-sleep 0.2
-
-printf "\n"
-printf "${C}● Phase 2  Uploading Tampa Bay sensor data...${NC}\n"
-sleep 0.4
-printf "  ${D}10 sensors  |  columns: id, name, lat, lon${NC}\n"
-sleep 0.8
-printf "  ${Y}[ingest]${NC}   running...\n"
-sleep 1.2
-printf "  ${G}[ingest]${NC}   completed ✓\n"
-sleep 0.3
-printf "${G}  ✓ Sensor table:${NC}  ${W}sensor_locations${NC}  (10 rows)\n"
-sleep 0.3
-
-printf "\n"
-printf "${C}● Phase 3  Running GPU geospatial queries...${NC}\n"
-sleep 0.3
-
-printf "\n"
-printf "  ${B}Query 1${NC}  Pairwise Distance  ${D}ST_DISTANCE()${NC}\n"
-sleep 1.0
-printf "  ┌────────────┬────────────┬────────────┐\n"
-printf "  │ sensor_a   │ sensor_b   │ distance_m │\n"
-printf "  ├────────────┼────────────┼────────────┤\n"
-printf "  │ Sensor A   │ Sensor B   │      2,847 │\n"
-printf "  │ Sensor A   │ Sensor J   │      4,213 │\n"
-printf "  │ Sensor B   │ Sensor J   │      5,091 │\n"
-printf "  └────────────┴────────────┴────────────┘\n"
-sleep 0.5
-
-printf "\n"
-printf "  ${B}Query 2${NC}  Point-in-Polygon  ${D}ST_CONTAINS()${NC}\n"
-sleep 0.9
-printf "  Zone: Downtown Tampa bounding polygon\n"
-printf "  ${G}  2 sensors inside zone:${NC}  Sensor A, Sensor B\n"
-sleep 0.5
-
-printf "\n"
-printf "  ${B}Query 3${NC}  Track Line  ${D}ST_MAKELINE()${NC}\n"
-sleep 0.9
-printf "  LINESTRING(-82.4572 27.9506, -82.4398 27.9659, ...)\n"
-printf "  ${G}  Track complete:${NC}  10 points\n"
-
-printf "\n"
-printf "${W}RESULT:${NC}\n"
-printf "  ${D}namespace_id:${NC}    a3f2c1d9-4b2e-4a7c-9f1d-8e3b0c2d6a5f\n"
-printf "  ${D}sensor_table:${NC}    sensor_locations  (10 rows)\n"
-printf "  ${D}phases_run:${NC}      [0, 1, 2, 3]\n"
-printf "\n"
-printf "  ${D}distance_query:${NC}\n"
-printf "    ${D}closest_pair:${NC}  Sensor A → Sensor B\n"
-printf "    ${D}min_distance:${NC}  2,847 m\n"
-printf "\n"
-printf "  ${D}point_in_polygon:${NC}\n"
-printf "    ${D}in_zone_count:${NC}   2\n"
-printf "    ${D}in_zone_sensors:${NC} [Sensor A, Sensor B]\n"
-printf "\n"
-printf "  ${D}track:${NC}\n"
-printf "    ${D}point_count:${NC}  10\n"
-printf "\n"

package/assets/demos/toolbelt-start-demo.sh

@@ -1,66 +0,0 @@
-#!/usr/bin/env bash
-# Simulates the toolbelt-start skill running inside Claude Code
-
-G='\033[0;32m'   # green
-B='\033[0;34m'   # blue
-Y='\033[1;33m'   # yellow
-C='\033[0;36m'   # cyan
-D='\033[2;37m'   # dim
-W='\033[1;37m'   # bold white
-NC='\033[0m'
-
-printf "\n"
-printf "${D}> /toolbelt-start document_url=https://docs.toolbelt.ai/intro.pdf question=\"What is Toolbelt?\"${NC}\n"
-sleep 0.8
-
-printf "\n"
-printf "${C}● Phase 0  Verifying Toolbelt MCP connection...${NC}\n"
-sleep 0.6
-printf "${G}  ✓ Connected${NC}  |  namespace: demo-workspace  |  1 namespace found\n"
-sleep 0.3
-
-printf "\n"
-printf "${C}● Phase 1  Resolving namespace...${NC}\n"
-sleep 0.4
-printf "${G}  ✓ Using${NC}  demo-workspace  (a3f2c1d9-4b2e-4a7c-9f1d-8e3b0c2d6a5f)\n"
-sleep 0.2
-
-printf "\n"
-printf "${C}● Phase 2  Inspecting current state...${NC}\n"
-sleep 0.5
-printf "${G}  ✓ Context loaded${NC}  |  0 tables  |  0 vector collections\n"
-sleep 0.2
-
-printf "\n"
-printf "${C}● Phase 3  Uploading document...${NC}\n"
-sleep 0.4
-printf "  ${D}file:${NC} intro.pdf  ${D}source:${NC} docs.toolbelt.ai\n"
-sleep 0.8
-printf "  ${Y}[ingest]${NC}   running...\n"
-sleep 1.2
-printf "  ${G}[ingest]${NC}   completed ✓\n"
-sleep 0.4
-printf "  ${Y}[semantic]${NC} running...\n"
-sleep 1.6
-printf "  ${G}[semantic]${NC} completed ✓\n"
-sleep 0.3
-printf "${G}  ✓ Ingested${NC}  →  table: ${W}intro_pdf${NC}\n"
-sleep 0.3
-
-printf "\n"
-printf "${C}● Phase 5  Answering question...${NC}\n"
-sleep 0.4
-printf "  ${D}toolbelt_search  question=\"What is Toolbelt?\"${NC}\n"
-sleep 1.2
-printf "${G}  ✓ Done${NC}\n"
-
-printf "\n"
-printf "${W}RESULT:${NC}\n"
-printf "  ${D}namespace_id:${NC}   a3f2c1d9-4b2e-4a7c-9f1d-8e3b0c2d6a5f\n"
-printf "  ${D}phases_run:${NC}     [0, 1, 2, 3, 5]\n"
-printf "  ${D}document_table:${NC} intro_pdf\n"
-printf "  ${D}answer:${NC}         Toolbelt is a GPU-accelerated data workspace\n"
-printf "                  providing SQL, vector search, knowledge graphs,\n"
-printf "                  geospatial queries, and Kafka streaming via MCP.\n"
-printf "  ${D}sources:${NC}        [intro.pdf]\n"
-printf "\n"