@realtimex/sdk 1.7.21 → 2.0.1

package/skills/realtimex-moderator-sdk/references/browser.md

@@ -1,27 +0,0 @@
-# RealTimeX Browser
-
-> Generated workflow guide · SDK **1.7.21** · 2026-05-18
-
-Use this for managed RealTimeX Browser sessions and tabs.
-
-Correct namespace:
-
-```js
-sdk.desktopBrowser
-```
-
-Preferred flow:
-1. Create or get a named browser session.
-2. Read its `remoteDebugPort`.
-3. Use the `agent-browser` skill against that CDP port for page interaction.
-
-```js
-await sdk.desktopBrowser.createSession({ sessionName: "docs-research" });
-await sdk.desktopBrowser.createTab({
-  sessionName: "docs-research",
-  url: "https://example.com"
-});
-const session = await sdk.desktopBrowser.getSession("docs-research");
-```
-
-Avoid mutating reserved `acp-*` browser sessions unless the user explicitly asks for internal ACP browser flows.

package/skills/realtimex-moderator-sdk/references/channels.md

@@ -1,189 +0,0 @@
-# External Chat Channels
-
-> Generated workflow guide · SDK **1.7.21** · 2026-05-18
-
-Use this for Telegram, Zalo, WhatsApp, Discord, Slack, and other chat channel setup.
-
-Required LocalApp permission for `x-app-id` mode:
-
-```js
-permissions: ["channels.manage"]
-```
-
-Main namespace:
-
-```js
-sdk.v1.channels
-```
-
-## Agent Setup Rules
-
-- Never print bot tokens, credentials, QR auth state, or full config values back to chat.
-- Do not guess workspace or thread. Load `skills get workspaces` if context is missing.
-- Ask the user to complete provider-side steps that cannot be automated, such as BotFather setup or QR scanning.
-- Prefer creating plugins disabled first, then start only after credentials/login/policies are ready.
-- Use `settings.thread_id` only when the user explicitly wants messages routed to a specific thread.
-- Use `agentWhitelist: ["*"]` only when the user wants any mentioned/available agent allowed. Otherwise ask which agents should be allowed.
-
-## Decision Tree
-
-- User says Telegram: ask for workspace and Telegram bot token. If they do not have a token, tell them to create one in BotFather first.
-- User says WhatsApp: create a `whatsapp` plugin with empty config, start QR login, ask the user to scan, poll state, configure policies, then start.
-- User says Zalo personal: create a `zalo_personal` plugin with empty config, start QR login, ask the user to scan, poll state, configure policies, then start.
-- User asks to restrict access: use pairing codes or provider policies.
-- User asks to allow anyone: set permissive policies only after confirming the security tradeoff.
-- User is unsure which provider: list supported values: `telegram`, `slack`, `discord`, `zalo`, `zalo_personal`, `whatsapp`.
-
-## Common Flow
-
-1. Resolve workspace and optional target thread.
-2. Identify provider and collect only the needed credential or QR action.
-3. For token providers, call `pluginsTest(...)` before creating when supported.
-4. Create plugin with `createPlugin(...)` and `enabled: false`.
-5. For QR providers, call `pluginsQrLoginStart(...)`, ask the user to scan, then poll `getState(...)` until connected or failed.
-6. Configure policies with `pluginsPolicies(...)` where relevant.
-7. Start with `pluginsStart(...)`.
-8. Verify `getStatus()` and ask the user to send a first message from the external platform.
-
-## Telegram Bot
-
-Provider-side step: the user must create a Telegram bot with BotFather and provide the bot token.
-
-Test credentials:
-
-```js
-await sdk.v1.channels.pluginsTest({
-  plugin_type: "telegram",
-  config: { botToken: process.env.TELEGRAM_BOT_TOKEN }
-});
-```
-
-Create plugin:
-
-```js
-const created = await sdk.v1.channels.createPlugin({
-  workspace_id: 1,
-  plugin_type: "telegram",
-  name: "Support Telegram",
-  enabled: false,
-  config: { botToken: process.env.TELEGRAM_BOT_TOKEN },
-  settings: { thread_id: null, agentWhitelist: ["*"] }
-});
-await sdk.v1.channels.pluginsStart(String(created.plugin.id));
-```
-
-First-message check: ask the user to open Telegram, start the bot, and send a test message.
-
-## WhatsApp QR
-
-Provider-side step: the user must scan the QR code with WhatsApp.
-
-```js
-const created = await sdk.v1.channels.createPlugin({
-  workspace_id: 1,
-  plugin_type: "whatsapp",
-  name: "WhatsApp",
-  enabled: false,
-  config: {},
-  settings: { thread_id: null, agentWhitelist: ["*"] }
-});
-
-const id = String(created.plugin.id);
-await sdk.v1.channels.pluginsQrLoginStart(id, { force: false });
-
-// Poll until status.connected is true, status.status is error, or user cancels.
-const state = await sdk.v1.channels.getState(id);
-
-await sdk.v1.channels.pluginsPolicies(id, {
-  policies: {
-    dmPolicy: "pairing",
-    groupPolicy: "disabled",
-    selfChatMode: false
-  }
-});
-
-await sdk.v1.channels.pluginsStart(id);
-```
-
-Policy guidance:
-- `dmPolicy: "pairing"` is safer for private access.
-- Keep `groupPolicy: "disabled"` unless the user explicitly wants group chat support.
-- Enable `selfChatMode` only when the user understands the loop/testing behavior.
-
-## Zalo Personal QR
-
-Provider-side step: the user must scan the QR code with Zalo.
-
-```js
-const created = await sdk.v1.channels.createPlugin({
-  workspace_id: 1,
-  plugin_type: "zalo_personal",
-  name: "Zalo Personal",
-  enabled: false,
-  config: {},
-  settings: { thread_id: null, agentWhitelist: ["*"] }
-});
-
-const id = String(created.plugin.id);
-await sdk.v1.channels.pluginsQrLoginStart(id, { force: false });
-const state = await sdk.v1.channels.getState(id);
-
-await sdk.v1.channels.pluginsPolicies(id, {
-  policies: {
-    dmPolicy: "pairing",
-    groupPolicy: "disabled",
-    requireMention: false,
-    allowFrom: [],
-    groups: {}
-  }
-});
-
-await sdk.v1.channels.pluginsStart(id);
-```
-
-Directory helpers after login:
-
-```js
-await sdk.v1.channels.listDirectoryFriends(id);
-await sdk.v1.channels.listDirectoryGroups(id);
-```
-
-## Pairing And User Approval
-
-Use pairing when the channel should not allow every external user by default.
-
-```js
-const code = await sdk.v1.channels.pluginsPairingCodes(pluginId, {
-  platform_user_id: "external-user-id",
-  platform_username: "Customer Name"
-});
-
-await sdk.v1.channels.listPluginPairingCodes(pluginId);
-await sdk.v1.channels.pairingCodesApprove(String(code.code.id));
-await sdk.v1.channels.listPluginUsers(pluginId);
-await sdk.v1.channels.pluginsUsersAuthorization(pluginId, userId, { authorized: true });
-```
-
-## Status And Troubleshooting
-
-```js
-await sdk.v1.channels.listPlugins();
-await sdk.v1.channels.getStatus();
-await sdk.v1.channels.getConfig(pluginId);
-```
-
-- `Not authenticated`: QR login did not complete or credentials were cleared. Start QR login again.
-- Start fails: stop the plugin, test credentials if token-based, check `getStatus()`, then start again.
-- User cannot chat: check pairing codes, user authorization, `dmPolicy`, `groupPolicy`, and provider allow lists.
-- Messages route to the wrong place: check `workspace_id` and `settings.thread_id`.
-- Bot does not receive external messages: confirm provider-side webhook/session requirements and that the plugin is running.
-
-## Cleanup
-
-```js
-await sdk.v1.channels.pluginsStop(pluginId);
-await sdk.v1.channels.pluginsLogout(pluginId); // QR providers only
-await sdk.v1.channels.deletePlugin(pluginId);
-```
-
-Exact generated methods are in `api-reference/v1-channels.md`.

package/skills/realtimex-moderator-sdk/references/credentials.md

@@ -1,111 +0,0 @@
-# Credentials Reference
-
-Credentials are managed by the user in **Settings > Credentials**. Agents have read-only access via `sdk.credentials`.
-
-## Security Rules
-
-1. **Never** print credential values to stdout (they become tool results in chat history)
-2. **Never** include credential values in your response text
-3. **Never** write credential values to files or logs
-4. **Never** pass credential values as CLI arguments (visible in process list)
-5. **Always** consume credentials inside scripts — fetch, use, discard
-6. **Always** call `process.exit(0)` at the end of custom scripts (prevents hanging on open HTTP sockets)
-7. **Never** write helper scripts into the SKILL directory — use the working directory or system temp
-
-## Credential Types
-
-| Type | Payload | How to Apply |
-|------|---------|-------------|
-| `http_header` | `{ name, value }` | Set as HTTP header: `headers[payload.name] = payload.value` |
-| `query_auth` | `{ name, value }` | Append to URL: `?${payload.name}=${payload.value}` |
-| `basic_auth` | `{ username, password }` | Encode: `Authorization: Basic ${btoa(username + ":" + password)}` |
-| `env_var` | `{ name, value }` | Set in subprocess: `env[payload.name] = payload.value` |
-
-## Usage Patterns
-
-### List available credentials
-
-```bash
-node "$SKILL" credentials
-```
-
-Output: table of names and types (no values).
-
-### Important: credential payload structure
-
-`sdk.credentials.get(name)` returns `{ name, type, payload }`. The `payload` object contains structured fields — use them directly. **Never** extract raw values and construct headers manually.
-
-### Use an http_header credential in a script
-
-```javascript
-const { initSDK } = require('<SKILL_DIR>/scripts/lib/sdk-init');
-(async () => {
-  const { sdk } = await initSDK();
-  const cred = await sdk.credentials.get('github-token');
-  // cred.type === "http_header"
-  // cred.payload === { name: "Authorization", value: "Bearer ghp_xxx" }
-  // The value already includes the full header value — use as-is
-  const res = await fetch('https://api.github.com/user', {
-    headers: { [cred.payload.name]: cred.payload.value }
-  });
-  console.log('Status:', res.status); // Only non-sensitive output
-  process.exit(0);
-})();
-// WRONG: const token = cred.payload.value; headers.Authorization = `Bearer ${token}`
-// The value already contains "Bearer ..." — adding prefix again causes 401
-```
-
-### Use a basic_auth credential
-
-```javascript
-const { initSDK } = require('<SKILL_DIR>/scripts/lib/sdk-init');
-(async () => {
-  const { sdk } = await initSDK();
-  const cred = await sdk.credentials.get('registry-login');
-  const auth = Buffer.from(cred.payload.username + ':' + cred.payload.password).toString('base64');
-  const res = await fetch('https://registry.example.com/v2/_catalog', {
-    headers: { 'Authorization': 'Basic ' + auth }
-  });
-  console.log('Status:', res.status);
-  process.exit(0);
-})();
-```
-
-### Use an env_var credential with a subprocess
-
-```javascript
-const { initSDK } = require('<SKILL_DIR>/scripts/lib/sdk-init');
-const { execSync } = require('child_process');
-(async () => {
-  const { sdk } = await initSDK();
-  const cred = await sdk.credentials.get('aws-key');
-  // cred.type === "env_var"
-  // cred.payload === { name: "AWS_ACCESS_KEY_ID", value: "AKIA..." }
-  execSync('aws s3 ls', {
-    env: { ...process.env, [cred.payload.name]: cred.payload.value },
-    stdio: 'inherit'
-  });
-  process.exit(0);
-})();
-```
-
-### Error handling
-
-```javascript
-const { initSDK } = require('<SKILL_DIR>/scripts/lib/sdk-init');
-(async () => {
-  try {
-    const { sdk } = await initSDK();
-    const cred = await sdk.credentials.get('my-key');
-    // use cred...
-    process.exit(0);
-  } catch (err) {
-    if (err.message.includes('not found')) {
-      console.log('Credential not found. Ask the user to add it in Settings > Credentials.');
-    } else {
-      console.log('Error:', err.message);
-    }
-    process.exit(1);
-  }
-})();
-```

package/skills/realtimex-moderator-sdk/references/known-issues.md

@@ -1,237 +0,0 @@
-# Known Issues — Source-Detected
-
-> Auto-generated by `scripts/generate-skill.mjs` · SDK **1.7.21** · 2026-05-18
-
-Run `node scripts/generate-skill.mjs --force` after SDK source changes to refresh.
-
-**12/12 issues confirmed in current source.**
-
----
-
-## 1. `sdk.webhook.triggerAgent()` sends `event: "task.trigger"` — server expects `"trigger-agent"`
-
-**Status:** 🔴 DETECTED  
-**File:** `modules/webhook.ts`  
-**Evidence:** `Found `event: "task.trigger"` hardcoded in triggerAgent() body (line ~100)`
-
-**Correct usage:**
-```js
-// ❌ WRONG — SDK sends 'task.trigger' but server enum only accepts 'trigger-agent'
-await sdk.webhook.triggerAgent({ auto_run: true, agent_name: 'agent', ... });
-
-// ✅ CORRECT — raw fetch with server-expected event string
-const resp = await fetch('http://localhost:3001/webhooks/realtimex', {
-  method: 'POST',
-  headers: { 'Content-Type': 'application/json', Authorization: `Bearer ${apiKey}` },
-  body: JSON.stringify({
-    event: 'trigger-agent',
-    payload: { auto_run: true, agent_name: 'document2speech', workspace_slug: 'sdfsdf', prompt: '...' },
-  }),
-});
-```
-
----
-
-## 2. `sdk.task.start/complete/fail` take positional `(taskUuid, ...)` — NOT `{ task_uuid }` object
-
-**Status:** 🔴 DETECTED  
-**File:** `modules/task.ts`  
-**Evidence:** `start(taskUuid: string, ...) | complete(taskUuid, result, ...) | fail(taskUuid, error: string, ...)`
-
-**Correct usage:**
-```js
-// ❌ WRONG
-await sdk.task.start({ task_uuid: uuid });
-await sdk.task.complete({ task_uuid: uuid, result: { out: 'done' } });
-
-// ✅ CORRECT
-await sdk.task.start(uuid);
-await sdk.task.complete(uuid, { out: 'done' });
-await sdk.task.fail(uuid, 'error message');
-await sdk.task.start(uuid, { machineId: 'machine-01' });
-```
-
----
-
-## 3. `sdk.activities.list()` returns `Activity[]` directly — NOT `{ activities: [...] }`
-
-**Status:** 🔴 DETECTED  
-**File:** `modules/activities.ts`  
-**Evidence:** `Return type is Promise<Activity[]>`
-
-**Correct usage:**
-```js
-// ❌ WRONG
-const { activities } = await sdk.activities.list();
-
-// ✅ CORRECT
-const activities = await sdk.activities.list();
-for (const a of activities) { ... }
-```
-
----
-
-## 4. `sdk.llm.chat()` response is `res.response?.content` — NOT `choices[0].message.content`
-
-**Status:** 🔴 DETECTED  
-**File:** `modules/llm.ts`  
-**Evidence:** `ChatResponse shape: { success, response?: { content, model, metrics } }`
-
-**Correct usage:**
-```js
-// ❌ WRONG (OpenAI style)
-console.log(res.choices[0].message.content);
-
-// ✅ CORRECT
-const res = await sdk.llm.chat([{ role: 'user', content: 'Hello' }]);
-console.log(res.response?.content);
-```
-
----
-
-## 5. `sdk.llm.chatStream()` yields `chunk.textResponse` — NOT `choices[0].delta.content`
-
-**Status:** 🔴 DETECTED  
-**File:** `modules/llm.ts`  
-**Evidence:** `StreamChunk interface has textResponse?: string`
-
-**Correct usage:**
-```js
-// ❌ WRONG
-chunk.choices?.[0]?.delta?.content
-
-// ✅ CORRECT
-for await (const chunk of sdk.llm.chatStream(messages)) {
-  if (chunk.textResponse) process.stdout.write(chunk.textResponse);
-  if (chunk.close) break;
-}
-```
-
----
-
-## 6. `sdk.llm.embedAndStore()` takes `{ texts: string[], documentId?, workspaceId?, ... }` — NOT `(text, options)`
-
-**Status:** 🔴 DETECTED  
-**File:** `modules/llm.ts`  
-**Evidence:** `embedAndStore(params: { texts: string[]; documentId?; workspaceId?; ... })`
-
-**Correct usage:**
-```js
-// ❌ WRONG
-await sdk.llm.embedAndStore('text content', { id: 'doc-1' });
-
-// ✅ CORRECT
-await sdk.llm.embedAndStore({ texts: ['chunk 1', 'chunk 2'], documentId: 'doc-1', workspaceId: 'ws-123' });
-```
-
----
-
-## 7. `sdk.llm.vectors.query()` takes a raw `number[]` embedding — NOT a text string
-
-**Status:** 🔴 DETECTED  
-**File:** `modules/llm.ts`  
-**Evidence:** `query(vector: number[], options?)`
-
-**Correct usage:**
-```js
-// ❌ WRONG
-await sdk.llm.vectors.query('search text');
-
-// ✅ CORRECT — raw vector
-const { embeddings } = await sdk.llm.embed('search text');
-await sdk.llm.vectors.query(embeddings[0], { topK: 5 });
-
-// ✅ EASIER — high-level helper
-const hits = await sdk.llm.search('search text', { topK: 5 });
-```
-
----
-
-## 8. `sdk.llm.vectors.delete()` requires `{ deleteAll: true }` — delete-by-ID not supported
-
-**Status:** 🔴 DETECTED  
-**File:** `modules/llm.ts`  
-**Evidence:** `VectorDeleteOptions: { workspaceId?, deleteAll: true } (literal true)`
-
-**Correct usage:**
-```js
-// ❌ WRONG
-await sdk.llm.vectors.delete(['id-1', 'id-2']);
-
-// ✅ CORRECT
-await sdk.llm.vectors.delete({ deleteAll: true, workspaceId: 'ws-123' });
-```
-
----
-
-## 9. `sdk.mcp.getServers()` takes a plain string — NOT `{ provider: "all" }`
-
-**Status:** 🔴 DETECTED  
-**File:** `modules/mcp.ts`  
-**Evidence:** `getServers(provider: 'local' | 'remote' | 'all' = 'all')`
-
-**Correct usage:**
-```js
-// ❌ WRONG
-await sdk.mcp.getServers({ provider: 'all' });
-
-// ✅ CORRECT
-await sdk.mcp.getServers('all');
-await sdk.mcp.getServers();     // defaults to 'all'
-```
-
----
-
-## 10. `getAgents/getWorkspaces/getThreads/getTask` live on `sdk.api.*` — NOT directly on `sdk.*`
-
-**Status:** 🔴 DETECTED  
-**File:** `index.ts`  
-**Evidence:** `this.api = new ApiModule(...) — separate sub-module in constructor`
-
-**Correct usage:**
-```js
-// ❌ WRONG
-await sdk.getAgents();
-await sdk.getWorkspaces();
-
-// ✅ CORRECT
-await sdk.api.getAgents();
-await sdk.api.getWorkspaces();
-await sdk.api.getThreads('workspace-slug');
-await sdk.api.getTask('task-uuid');
-```
-
----
-
-## 11. ACP `streamChat` uses named SSE (`event:` line); `text_delta.data.type === "thinking"` = internal reasoning
-
-**Status:** 🔴 DETECTED  
-**File:** `modules/acpAgent.ts`  
-**Evidence:** `Uses parseNamedSSEStream() — reads both event: and data: SSE lines per spec`
-
-**Correct usage:**
-```js
-for await (const event of sdk.acpAgent.streamChat(key, message)) {
-  if (event.type === 'text_delta') {
-    if (event.data.type !== 'thinking') {   // skip internal reasoning
-      process.stdout.write(String(event.data.text ?? ''));
-    }
-  }
-}
-```
-
----
-
-## 12. ACP sessions stall without `approvalPolicy: "approve-all"` when tools need permission
-
-**Status:** 🔴 DETECTED  
-**File:** `modules/acpAgent.ts`  
-**Evidence:** `AcpSessionOptions.approvalPolicy?: 'approve-all' | 'approve-reads' | 'deny-all'`
-
-**Correct usage:**
-```js
-// ✅ Always set for autonomous scripts
-await sdk.acpAgent.createSession({ agent_id: 'qwen', approvalPolicy: 'approve-all', cwd });
-```
-
----

package/skills/realtimex-moderator-sdk/references/llm.md

@@ -1,13 +0,0 @@
-# LLM And Vector Store
-
-> Generated workflow guide · SDK **1.7.21** · 2026-05-18
-
-Use `sdk.llm` for chat, streaming, embeddings, and vector helpers.
-
-Common permissions: `llm.chat`, `llm.embed`, `llm.providers`, `vectors.read`, `vectors.write`.
-
-```js
-await sdk.llm.chat([{ role: "user", content: "Hello" }]);
-await sdk.llm.embed(["text to embed"]);
-await sdk.llm.vectors.query(vector, { workspaceId });
-```

package/skills/realtimex-moderator-sdk/references/mcp.md

@@ -1,13 +0,0 @@
-# MCP
-
-> Generated workflow guide · SDK **1.7.21** · 2026-05-18
-
-Use `sdk.mcp` to list MCP servers, list tools, and execute tools.
-
-Required permissions: `mcp.servers` and `mcp.tools`.
-
-```js
-await sdk.mcp.getServers();
-await sdk.mcp.getTools(serverName);
-await sdk.mcp.executeTool(serverName, toolName, args);
-```

package/skills/realtimex-moderator-sdk/references/permissions.md

@@ -1,30 +0,0 @@
-# Permissions
-
-> Generated workflow guide · SDK **1.7.21** · 2026-05-18
-
-LocalApps using `x-app-id` must request permissions before calling protected SDK routes.
-
-| Permission | Use For |
-|---|---|
-| `api.agents` | List agents |
-| `api.workspaces` | List workspaces |
-| `api.threads` | List workspace threads |
-| `api.task` | Read task status |
-| `activities.read` | Read activities |
-| `activities.write` | Create/update/delete activities |
-| `llm.chat` | LLM chat and streaming chat |
-| `llm.embed` | Generate embeddings |
-| `llm.providers` | List LLM/embed providers |
-| `vectors.read` | Query/list vector stores |
-| `vectors.write` | Upsert/delete vectors |
-| `mcp.servers` | List MCP servers |
-| `mcp.tools` | List/execute MCP tools |
-| `acp.agent` | ACP agent sessions |
-| `desktop.runtime-sessions` | Visible Electron terminal sessions |
-| `desktop.browser` | RealTimeX Browser sessions and tabs |
-| `channels.manage` | External chat channel setup and administration |
-| `tts.generate` | Text-to-speech |
-| `stt.listen` | Speech-to-text |
-| `credentials.read` | Read stored credentials |
-
-API-key dev mode has wildcard access.

package/skills/realtimex-moderator-sdk/references/quickstart.md

@@ -1,16 +0,0 @@
-# Quickstart
-
-> Generated workflow guide · SDK **1.7.21** · 2026-05-18
-
-Use this when starting any SDK task.
-
-```js
-const { initSDK } = require('<SKILL_DIR>/scripts/lib/sdk-init');
-const { sdk, context } = await initSDK();
-```
-
-Rules:
-- Use the working directory or system temp for helper scripts, never the skill directory.
-- Exit scripts explicitly with `process.exit(0)` or `process.exit(1)`.
-- Check `context.workspaceSlug` and `context.threadSlug` before asking the user.
-- For exact signatures, open `references/api-reference/index.md`.

package/skills/realtimex-moderator-sdk/references/terminal-sessions.md

@@ -1,34 +0,0 @@
-# Desktop Terminal Sessions
-
-> Generated workflow guide · SDK **1.7.21** · 2026-05-18
-
-Use this for visible Electron terminal sessions.
-
-Correct namespace:
-
-```js
-sdk.desktopRuntimeSessions
-```
-
-Do not use ACP for visible terminals unless the user explicitly asks for headless ACP.
-
-Examples:
-
-```js
-await sdk.desktopRuntimeSessions.launchTerminalCliAgent({
-  workspaceSlug,
-  threadSlug,
-  agentName: "claude",
-  providerId: "claude-cli",
-  presentationMode: "panel",
-  message: "what is current working dir"
-});
-
-await sdk.desktopRuntimeSessions.launchTerminalShell({
-  workspaceSlug,
-  threadSlug,
-  presentationMode: "panel",
-  initialCommand: "pwd",
-  initialCommandMode: "direct"
-});
-```

package/skills/realtimex-moderator-sdk/references/workspaces.md

@@ -1,20 +0,0 @@
-# Workspaces And Threads
-
-> Generated workflow guide · SDK **1.7.21** · 2026-05-18
-
-Use this before any task that needs workspace/thread context.
-
-Priority order:
-1. Explicit user-provided workspace/thread.
-2. `context.workspaceSlug` / `context.threadSlug` from `initSDK()`.
-3. `RTX_WORKSPACE_SLUG` / `RTX_THREAD_SLUG` in spawned sessions.
-4. List workspaces and threads, then ask only if ambiguous.
-
-Useful calls:
-
-```js
-await sdk.api.getWorkspaces();
-await sdk.api.getThreads(workspaceSlug);
-await sdk.v1.workspace.listWorkspaces();
-await sdk.v1.thread.listWorkspaceThreads(slug);
-```