zubo 0.1.21 → 0.1.23

package/.playwright-mcp/console-2026-02-16T19-06-21-167Z.log

@@ -0,0 +1,31 @@
+[  127722ms] [ERROR] Failed to load resource: net::ERR_CONNECTION_REFUSED @ http://localhost:57108/api/dashboard/memory:0
+[  127722ms] [ERROR] Failed to load resource: net::ERR_CONNECTION_REFUSED @ http://localhost:57108/api/dashboard/memory/recent:0
+[  127722ms] [WARNING] Dashboard API request failed TypeError: Failed to fetch
+    at api (http://localhost:57108/:2291:10)
+    at loadRecentMemories (http://localhost:57108/:2795:3)
+    at loadMemory (http://localhost:57108/:2756:5)
+    at showPanel (http://localhost:57108/:2089:26)
+    at HTMLAnchorElement.onclick (http://localhost:57108/#status:929:53) @ http://localhost:57108/:2796
+[  127754ms] TypeError: Failed to fetch
+    at api (http://localhost:57108/:2291:10)
+    at loadMemory (http://localhost:57108/:2750:3)
+    at showPanel (http://localhost:57108/:2089:26)
+    at HTMLAnchorElement.onclick (http://localhost:57108/#status:929:53)
+[  127724ms] [ERROR] Failed to load resource: net::ERR_CONNECTION_REFUSED @ http://localhost:57108/api/dashboard/memory:0
+[  127754ms] TypeError: Failed to fetch
+    at api (http://localhost:57108/:2291:10)
+    at loadMemory (http://localhost:57108/:2750:3)
+    at showPanel (http://localhost:57108/:2089:26)
+    at routeFromHash (http://localhost:57108/:2277:3)
+[  143412ms] [ERROR] Failed to load resource: net::ERR_CONNECTION_REFUSED @ http://localhost:57108/api/dashboard/skills:0
+[  143463ms] TypeError: Failed to fetch
+    at api (http://localhost:57108/:2291:10)
+    at loadSkills (http://localhost:57108/:2818:3)
+    at showPanel (http://localhost:57108/:2090:26)
+    at HTMLAnchorElement.onclick (http://localhost:57108/#status:932:53)
+[  143412ms] [ERROR] Failed to load resource: net::ERR_CONNECTION_REFUSED @ http://localhost:57108/api/dashboard/skills:0
+[  143463ms] TypeError: Failed to fetch
+    at api (http://localhost:57108/:2291:10)
+    at loadSkills (http://localhost:57108/:2818:3)
+    at showPanel (http://localhost:57108/:2090:26)
+    at routeFromHash (http://localhost:57108/:2277:3)

package/README.md

@@ -35,7 +35,8 @@
 - **Workflows** — Multi-agent pipelines with delegation
 - **Natural language scheduling** — "Every weekday at 9am" just works. Cron jobs, heartbeat, proactive tasks.
 - **Voice** — Speech-to-text (Whisper) and text-to-speech (OpenAI, ElevenLabs)
-- **Dashboard** — Built-in web UI with analytics, memory management, and settings
+- **Personal tools** — Todos, notes, preferences, topics, and follow-ups — all manageable from the dashboard or via chat
+- **Dashboard** — Built-in web UI with analytics, memory management, personal tools, and settings
 - **Document ingestion** — Upload PDF, DOCX, TXT, CSV, JSON, and more
 - **Budget controls** — Daily/monthly spending limits with per-model cost tracking
 - **100% local** — SQLite database, local vector store. Your data never leaves your machine.

package/docs/ROADMAP.md

@@ -439,52 +439,15 @@ Shows:
 
 ---
 
-## Implementation Priority
-
-What to build first, based on impact vs. effort:
-
-```
-HIGH IMPACT, LOW EFFORT (do first)
-├── Phase 1.1  OpenAI-compatible provider        ★★★★★
-├── Phase 1.2  Provider factory + config          ★★★★★
-├── Phase 1.5  `bun run model` command            ★★★★
-├── Phase 7.1  TUI / `bun run chat`              ★★★★
-├── Phase 4.3  Cost + token tracking              ★★★★
-└── Phase 5.2  Agent-managed cron tools           ★★★★
-
-HIGH IMPACT, MEDIUM EFFORT (do second)
-├── Phase 1.3  Failover wrapper                   ★★★★
-├── Phase 3.1  Skill format + loader              ★★★★
-├── Phase 3.3  web-search + url-fetch skills      ★★★★
-├── Phase 2.4  WebChat (local HTTP UI)            ★★★★
-├── Phase 4.1  Streaming responses                ★★★
-└── Phase 4.4  Context window awareness           ★★★
-
-MEDIUM IMPACT, MEDIUM EFFORT (do third)
-├── Phase 2.2  Discord channel                    ★★★
-├── Phase 2.3  WhatsApp channel                   ★★★
-├── Phase 5.1  Webhook inbound                    ★★★
-├── Phase 6.1  Tool approval / permissions        ★★★
-└── Phase 5.3  Daily digest                       ★★★
-
-NICE TO HAVE (do when stable)
-├── Phase 1.4  Ollama auto-discovery              ★★
-├── Phase 4.2  Extended thinking                  ★★
-├── Phase 6.2  Message pairing / auth             ★★
-└── Phase 7.2  Web dashboard                      ★★
-```
-
----
-
-## Next Concrete Step
-
-**Phase 1.1 + 1.2**: Multi-LLM support. One session of work:
-
-1. Create `src/llm/openai-compat.ts` — OpenAI-compatible provider
-2. Create `src/llm/factory.ts` — Provider factory
-3. Update `src/config/schema.ts` — New provider config fields
-4. Update `src/start.ts` — Use factory
-5. Update `src/setup.ts` — Provider selection in wizard
-6. Test with Ollama locally
-
-This single change makes Zubo usable with any model, which unlocks everything else.
+## Completed
+
+Most phases above are now implemented:
+
+- Phase 1 (Multi-LLM) — 11+ providers, failover, smart routing, CLI providers (Codex, Claude Code)
+- Phase 2 (Channels) — Telegram, Discord, Slack, WhatsApp, Signal, Email, WebChat
+- Phase 3 (Skills) — Skill loader, sandboxed execution, community registry, MCP support
+- Phase 4 (Agent) — Streaming, cost tracking, context window awareness, knowledge graph
+- Phase 5 (Automation) — Webhooks, agent-managed cron, daily digests, follow-ups
+- Phase 6 (Security) — Tool permissions (auto/confirm/deny), confirmation tokens, API key auth
+- Phase 7 (Interfaces) — Web dashboard with analytics, memory, skills, settings, and personal tools (todos, notes, preferences, topics, follow-ups)
+- Personal features — Todos, notes, preferences, topics, follow-ups with full CRUD in the dashboard

package/migrations/024_personal_features.sql

@@ -0,0 +1,96 @@
+-- Personal agent features: todos, notes, preferences, topics, follow-ups
+
+CREATE TABLE IF NOT EXISTS todos (
+  id INTEGER PRIMARY KEY AUTOINCREMENT,
+  title TEXT NOT NULL,
+  description TEXT,
+  priority TEXT NOT NULL DEFAULT 'medium' CHECK(priority IN ('low', 'medium', 'high', 'urgent')),
+  status TEXT NOT NULL DEFAULT 'pending' CHECK(status IN ('pending', 'in_progress', 'done')),
+  due_date TEXT,
+  tags TEXT NOT NULL DEFAULT '[]',
+  created_at TEXT NOT NULL DEFAULT (datetime('now')),
+  completed_at TEXT
+);
+
+CREATE INDEX IF NOT EXISTS idx_todos_status ON todos(status);
+CREATE INDEX IF NOT EXISTS idx_todos_priority ON todos(priority);
+CREATE INDEX IF NOT EXISTS idx_todos_due ON todos(due_date);
+
+CREATE TABLE IF NOT EXISTS notes (
+  id INTEGER PRIMARY KEY AUTOINCREMENT,
+  title TEXT NOT NULL,
+  content TEXT NOT NULL,
+  tags TEXT NOT NULL DEFAULT '[]',
+  pinned INTEGER NOT NULL DEFAULT 0,
+  created_at TEXT NOT NULL DEFAULT (datetime('now')),
+  updated_at TEXT NOT NULL DEFAULT (datetime('now'))
+);
+
+CREATE VIRTUAL TABLE IF NOT EXISTS notes_fts USING fts5(
+  title,
+  content,
+  tags,
+  content='notes',
+  content_rowid='id'
+);
+
+CREATE TRIGGER IF NOT EXISTS notes_ai AFTER INSERT ON notes BEGIN
+  INSERT INTO notes_fts(rowid, title, content, tags)
+  VALUES (new.id, new.title, new.content, new.tags);
+END;
+
+CREATE TRIGGER IF NOT EXISTS notes_ad AFTER DELETE ON notes BEGIN
+  INSERT INTO notes_fts(notes_fts, rowid, title, content, tags)
+  VALUES ('delete', old.id, old.title, old.content, old.tags);
+END;
+
+CREATE TRIGGER IF NOT EXISTS notes_au AFTER UPDATE ON notes BEGIN
+  INSERT INTO notes_fts(notes_fts, rowid, title, content, tags)
+  VALUES ('delete', old.id, old.title, old.content, old.tags);
+  INSERT INTO notes_fts(rowid, title, content, tags)
+  VALUES (new.id, new.title, new.content, new.tags);
+END;
+
+-- Upgrade user_preferences from migration 008 (simple key/value)
+-- to the expanded schema with category, confidence, source
+CREATE TABLE IF NOT EXISTS _user_preferences_new (
+  id INTEGER PRIMARY KEY AUTOINCREMENT,
+  category TEXT NOT NULL DEFAULT 'general',
+  key TEXT NOT NULL,
+  value TEXT NOT NULL,
+  confidence REAL NOT NULL DEFAULT 0.8,
+  source TEXT NOT NULL DEFAULT 'inferred',
+  created_at TEXT NOT NULL DEFAULT (datetime('now')),
+  updated_at TEXT NOT NULL DEFAULT (datetime('now')),
+  UNIQUE(category, key)
+);
+
+INSERT OR IGNORE INTO _user_preferences_new (key, value, updated_at)
+  SELECT key, value, updated_at FROM user_preferences;
+
+DROP TABLE IF EXISTS user_preferences;
+
+ALTER TABLE _user_preferences_new RENAME TO user_preferences;
+
+CREATE INDEX IF NOT EXISTS idx_prefs_category ON user_preferences(category);
+
+CREATE TABLE IF NOT EXISTS conversation_topics (
+  id INTEGER PRIMARY KEY AUTOINCREMENT,
+  name TEXT NOT NULL UNIQUE,
+  description TEXT,
+  status TEXT NOT NULL DEFAULT 'active' CHECK(status IN ('active', 'archived')),
+  last_message_at TEXT NOT NULL DEFAULT (datetime('now')),
+  created_at TEXT NOT NULL DEFAULT (datetime('now'))
+);
+
+CREATE TABLE IF NOT EXISTS follow_ups (
+  id INTEGER PRIMARY KEY AUTOINCREMENT,
+  context TEXT NOT NULL,
+  message TEXT NOT NULL,
+  follow_up_at TEXT NOT NULL,
+  status TEXT NOT NULL DEFAULT 'pending' CHECK(status IN ('pending', 'sent', 'cancelled')),
+  created_at TEXT NOT NULL DEFAULT (datetime('now'))
+);
+
+CREATE INDEX IF NOT EXISTS idx_followups_status ON follow_ups(status);
+CREATE INDEX IF NOT EXISTS idx_followups_time ON follow_ups(follow_up_at);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "zubo",
-  "version": "0.1.21",
+  "version": "0.1.23",
   "description": "Your AI agent that never forgets. Persistent memory, 25+ tools, 7 channels, 11+ LLM providers — runs entirely on your machine.",
   "license": "MIT",
   "author": "thomaskanze",

package/site/docs/index.html

@@ -61,6 +61,7 @@
           <a href="/docs/agents">Agents &amp; Workflows</a>
           <a href="/docs/memory">Memory System</a>
           <a href="/docs/skills">Skills</a>
+          <a href="/docs/skills#personal-features">Personal Features</a>
         </div>
       </div>
       <div class="docs-sidebar-section">
@@ -154,6 +155,15 @@
         <li><strong>Skill sandboxing</strong> &mdash; user-installed skills run in isolated subprocesses with configurable timeouts, preventing runaway code from affecting the main process.</li>
       </ul>
 
+      <h3>Personal Features</h3>
+      <ul>
+        <li><strong>Todos</strong> &mdash; add, list, complete, update, and remove tasks with priorities (<code>low</code> / <code>medium</code> / <code>high</code>), due dates, and tags. A full task manager inside your agent.</li>
+        <li><strong>Notes</strong> &mdash; save, search (FTS5), list, update, delete, and pin notes with tags. Fast full-text search across all your notes.</li>
+        <li><strong>Preferences</strong> &mdash; store user preferences as category/key/value pairs. Preferences are automatically injected into the system prompt so the agent always respects your settings.</li>
+        <li><strong>Topics</strong> &mdash; create, switch, list, and archive conversation topics. Each topic scopes a separate session, keeping context focused when you juggle multiple projects.</li>
+        <li><strong>Follow-ups</strong> &mdash; schedule proactive follow-up messages at a specific time or delay. Uses one-shot cron jobs that fire once and auto-delete.</li>
+      </ul>
+
       <h3>Agents</h3>
       <ul>
         <li><strong>Custom sub-agents</strong> with focused system prompts, restricted tool access, and specialized behavior. Define a &ldquo;code-reviewer&rdquo; agent that only has access to file and git tools, or a &ldquo;researcher&rdquo; agent that only has web search.</li>
@@ -640,6 +650,7 @@ zubo start --daemon</code></pre>
         <li><a href="/docs/agents"><strong>Agents &amp; Workflows</strong></a> &mdash; learn how to create custom sub-agents with focused system prompts and restricted tools, and build multi-agent pipelines with dependency resolution and parallel execution.</li>
         <li><a href="/docs/memory"><strong>Memory System</strong></a> &mdash; understand how semantic memory works, how to ingest documents, search your knowledge base, and tune the hybrid search weights.</li>
         <li><a href="/docs/skills"><strong>Skills</strong></a> &mdash; write, install, and share single-file TypeScript skills. Learn the skill manifest format, the sandbox environment, and the community registry.</li>
+        <li><a href="/docs/skills#personal-features"><strong>Personal Features</strong></a> &mdash; built-in tools for todos, notes, preferences, conversation topics, and scheduled follow-ups. Your personal productivity layer, stored locally.</li>
         <li><a href="/docs/channels"><strong>Channel Setup</strong></a> &mdash; step-by-step guides for configuring each channel: creating bots, obtaining tokens, setting up webhooks, and configuring allow-lists.</li>
         <li><a href="/docs/conversations"><strong>Conversation History</strong></a> &mdash; unified cross-channel history with FTS5 search, dashboard browsing, and API access for searching and analyzing past conversations across all channels.</li>
         <li><a href="/docs/webhooks"><strong>Webhooks</strong></a> &mdash; create webhook endpoints for GitHub, Stripe, CI/CD, and any external service. HMAC signature verification, prompt templates with <code>{{payload}}</code> substitution, and dashboard management.</li>

package/site/docs/skills.html

@@ -65,6 +65,7 @@
           <a href="/docs/agents">Agents &amp; Workflows</a>
           <a href="/docs/memory">Memory System</a>
           <a href="/docs/skills" class="active">Skills</a>
+          <a href="/docs/skills#personal-features">Personal Features</a>
         </div>
       </div>
       <div class="docs-sidebar-section">
@@ -295,10 +296,116 @@ zubo publish my_skill     # Publish your skill to the registry</code></pre>
           <tr><td><code>image_generate</code></td><td>Generate images using AI (DALL-E, etc.)</td><td>auto</td></tr>
           <tr><td><code>webhook_manage</code></td><td>Create, list, or remove incoming webhooks</td><td>auto</td></tr>
           <tr><td><code>oauth_manage</code></td><td>Manage OAuth provider connections</td><td>auto</td></tr>
+          <tr><td><code>todos</code></td><td>Add, list, complete, update, and remove tasks with priorities, due dates, and tags</td><td>auto</td></tr>
+          <tr><td><code>notes</code></td><td>Save, search (FTS5), list, update, delete, and pin notes with tags</td><td>auto</td></tr>
+          <tr><td><code>preferences</code></td><td>Set, get, list, and remove user preferences (injected into system prompt)</td><td>auto</td></tr>
+          <tr><td><code>topics</code></td><td>Create, switch, list, and archive conversation topics</td><td>auto</td></tr>
+          <tr><td><code>follow_ups</code></td><td>Schedule, list, and cancel proactive follow-up messages</td><td>auto</td></tr>
         </tbody>
       </table>
       <p>Tools marked <strong>confirm</strong> require user confirmation before execution. Tools marked <strong>auto</strong> run immediately without prompting.</p>
 
+      <h2 id="personal-features">Personal Features</h2>
+      <p>Built-in tools for personal productivity &mdash; todos, notes, preferences, conversation topics, and scheduled follow-ups. These run in the main process (not sandboxed) and store data in your local SQLite database.</p>
+
+      <h3><code>todos</code></h3>
+      <p>A full task manager built into your agent. Add tasks, set priorities and due dates, tag them for organization, and check them off when done.</p>
+      <table>
+        <thead>
+          <tr><th>Action</th><th>Description</th></tr>
+        </thead>
+        <tbody>
+          <tr><td><code>add</code></td><td>Create a new task with an optional priority (<code>low</code>, <code>medium</code>, <code>high</code>), due date, and tags.</td></tr>
+          <tr><td><code>list</code></td><td>List tasks, optionally filtered by status, priority, tag, or due date.</td></tr>
+          <tr><td><code>complete</code></td><td>Mark a task as done by ID.</td></tr>
+          <tr><td><code>update</code></td><td>Update a task's text, priority, due date, or tags.</td></tr>
+          <tr><td><code>remove</code></td><td>Delete a task by ID.</td></tr>
+        </tbody>
+      </table>
+      <p><strong>Key parameters:</strong> <code>text</code> (string), <code>priority</code> (<code>"low"</code> | <code>"medium"</code> | <code>"high"</code>), <code>due</code> (date string), <code>tags</code> (comma-separated), <code>id</code> (task ID for update/complete/remove).</p>
+      <p><strong>Example:</strong></p>
+<pre><code>"Add a todo: finish the API docs, high priority, due Friday, tag: work"
+"Show my todos tagged work"
+"Complete todo #3"</code></pre>
+
+      <h3><code>notes</code></h3>
+      <p>Save, search, and organize notes with full-text search (FTS5). Pin important notes so they surface first.</p>
+      <table>
+        <thead>
+          <tr><th>Action</th><th>Description</th></tr>
+        </thead>
+        <tbody>
+          <tr><td><code>save</code></td><td>Create a new note with optional title, tags, and pinned status.</td></tr>
+          <tr><td><code>search</code></td><td>Full-text search across all notes using FTS5.</td></tr>
+          <tr><td><code>list</code></td><td>List all notes, optionally filtered by tag or pinned status.</td></tr>
+          <tr><td><code>update</code></td><td>Update a note's content, title, tags, or pinned status.</td></tr>
+          <tr><td><code>delete</code></td><td>Delete a note by ID.</td></tr>
+          <tr><td><code>pin</code></td><td>Toggle the pinned status of a note.</td></tr>
+        </tbody>
+      </table>
+      <p><strong>Key parameters:</strong> <code>title</code> (string), <code>content</code> (string), <code>tags</code> (comma-separated), <code>pinned</code> (boolean), <code>query</code> (search string), <code>id</code> (note ID).</p>
+      <p><strong>Example:</strong></p>
+<pre><code>"Save a note: meeting with design team moved to Thursday 2pm, tag: meetings"
+"Search my notes for API design"
+"Pin note #5"</code></pre>
+
+      <h3><code>preferences</code></h3>
+      <p>Store user preferences as category/key/value pairs. Preferences are automatically injected into the system prompt so the agent always knows your settings.</p>
+      <table>
+        <thead>
+          <tr><th>Action</th><th>Description</th></tr>
+        </thead>
+        <tbody>
+          <tr><td><code>set</code></td><td>Set a preference value for a given category and key.</td></tr>
+          <tr><td><code>get</code></td><td>Retrieve a specific preference by category and key.</td></tr>
+          <tr><td><code>list</code></td><td>List all preferences, optionally filtered by category.</td></tr>
+          <tr><td><code>remove</code></td><td>Delete a preference by category and key.</td></tr>
+        </tbody>
+      </table>
+      <p><strong>Key parameters:</strong> <code>category</code> (string, e.g. <code>"display"</code>, <code>"communication"</code>, <code>"coding"</code>), <code>key</code> (string), <code>value</code> (string).</p>
+      <p><strong>Example:</strong></p>
+<pre><code>"Set my preference: coding/language = TypeScript"
+"What are my communication preferences?"
+"Remove preference coding/editor"</code></pre>
+
+      <h3><code>topics</code></h3>
+      <p>Organize conversations into named topics. Each topic scopes a separate session, so context stays focused and you can switch between projects easily.</p>
+      <table>
+        <thead>
+          <tr><th>Action</th><th>Description</th></tr>
+        </thead>
+        <tbody>
+          <tr><td><code>create</code></td><td>Create a new topic with a name and optional description.</td></tr>
+          <tr><td><code>switch</code></td><td>Switch the active conversation to a different topic.</td></tr>
+          <tr><td><code>list</code></td><td>List all topics, including archived ones.</td></tr>
+          <tr><td><code>archive</code></td><td>Archive a topic to keep things tidy without deleting history.</td></tr>
+        </tbody>
+      </table>
+      <p><strong>Key parameters:</strong> <code>name</code> (string), <code>description</code> (string), <code>id</code> (topic ID for switch/archive).</p>
+      <p><strong>Example:</strong></p>
+<pre><code>"Create a topic called 'Website Redesign'"
+"Switch to the API project topic"
+"List my topics"
+"Archive the old marketing topic"</code></pre>
+
+      <h3><code>follow_ups</code></h3>
+      <p>Schedule proactive follow-up messages that Zubo sends to you at a specified time. Uses one-shot cron jobs under the hood &mdash; the job fires once and auto-deletes.</p>
+      <table>
+        <thead>
+          <tr><th>Action</th><th>Description</th></tr>
+        </thead>
+        <tbody>
+          <tr><td><code>schedule</code></td><td>Schedule a follow-up message for a specific time or delay.</td></tr>
+          <tr><td><code>list</code></td><td>List all pending follow-ups.</td></tr>
+          <tr><td><code>cancel</code></td><td>Cancel a scheduled follow-up by ID.</td></tr>
+        </tbody>
+      </table>
+      <p><strong>Key parameters:</strong> <code>message</code> (string &mdash; what Zubo should say), <code>time</code> (date/time string or natural language like <code>"in 2 hours"</code>, <code>"tomorrow at 9am"</code>), <code>id</code> (follow-up ID for cancel).</p>
+      <p><strong>Example:</strong></p>
+<pre><code>"Follow up with me tomorrow at 10am about the deployment"
+"List my pending follow-ups"
+"Cancel follow-up #2"</code></pre>
+
       <h2>Legacy Format (SKILL.md)</h2>
       <p>Earlier versions of Zubo used a Markdown-based skill definition file called <code>SKILL.md</code>. This format is still fully supported for backward compatibility:</p>
 <pre><code># weather

package/site/index.html

@@ -300,7 +300,15 @@
           <p>Build your own tools in TypeScript and share them with the community. A sandboxed skill system with scoped secrets, hot-reload, and a public registry.</p>
         </div>
 
-        <div class="bento-card bento-lg tilt-card" data-feature="llm">
+        <div class="bento-card tilt-card" data-feature="personal">
+          <div class="bento-icon">
+            <svg width="28" height="28" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"><path d="M9 11l3 3L22 4"/><path d="M21 12v7a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V5a2 2 0 0 1 2-2h11"/></svg>
+          </div>
+          <h3>Personal Features</h3>
+          <p>Built-in todos, notes, preferences, conversation topics, and scheduled follow-ups. Your personal productivity layer, stored locally and always in context.</p>
+        </div>
+
+        <div class="bento-card tilt-card" data-feature="llm">
           <div class="bento-icon">
             <svg width="28" height="28" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"><path d="M21 16V8a2 2 0 0 0-1-1.73l-7-4a2 2 0 0 0-2 0l-7 4A2 2 0 0 0 3 8v8a2 2 0 0 0 1 1.73l7 4a2 2 0 0 0 2 0l7-4A2 2 0 0 0 21 16z"/><polyline points="7.5 4.21 12 6.81 16.5 4.21"/><polyline points="7.5 19.79 7.5 14.6 3 12"/><polyline points="21 12 16.5 14.6 16.5 19.79"/><polyline points="3.27 6.96 12 12.01 20.73 6.96"/><line x1="12" y1="22.08" x2="12" y2="12"/></svg>
           </div>

package/src/agent/context.ts

@@ -7,12 +7,12 @@ export interface AgentContext {
   messages: LlmMessage[];
 }
 
-export function assembleContext(
+export async function assembleContext(
   sessionId: string,
   maxTurns: number = 50,
   memories: string = ""
-): AgentContext {
+): Promise<AgentContext> {
   const messages = loadSession(sessionId, maxTurns);
-  const system = buildSystemPrompt(memories);
+  const system = await buildSystemPrompt(memories);
   return { system, messages };
 }

package/src/agent/delegate.ts

@@ -1,6 +1,7 @@
 import type { LlmProvider } from "../llm/provider";
 import { getAgentDefinition } from "./agents";
 import { agentLoop } from "./loop";
+import { getAllToolDefs } from "../tools/registry";
 import { searchMemory } from "../memory/engine";
 import { getDb } from "../db/connection";
 import { logger } from "../util/logger";
@@ -107,7 +108,11 @@ export async function delegateToAgent(
     "delegate", "delegate_task", "manage_agents",
     "config_update", "secret_set", "secret_delete", "manage_skills",
   ]);
-  const filteredTools = agent.tools.filter(
+  // Support wildcard "*" — gives the agent all non-forbidden tools dynamically
+  const baseTools = agent.tools.includes("*")
+    ? getAllToolDefs().map(t => t.name)
+    : agent.tools;
+  const filteredTools = baseTools.filter(
     (t) => !FORBIDDEN_DELEGATION_TOOLS.has(t)
   );
 
@@ -115,7 +120,7 @@ export async function delegateToAgent(
   try {
     const result = await agentLoop(llm, sessionId, task, {
       systemPromptOverride: systemPrompt,
-      allowedTools: filteredTools.length > 0 ? filteredTools : undefined,
+      allowedTools: filteredTools,
       maxRounds: 8,
       memories,
     });

package/src/agent/loop.ts

@@ -39,12 +39,12 @@ function resolveOptions(memoriesOrOptions: string | AgentLoopOptions): AgentLoop
     : memoriesOrOptions;
 }
 
-/** Detect simple greetings/chat that don't need tool definitions in context. */
+/** Detect standalone greetings that don't need tool definitions in context. */
 function looksConversational(text: string): boolean {
-  const t = text.trim().toLowerCase();
-  if (t.split(/\s+/).length > 8) return false; // longer messages likely need tools
-  const greetings = /^(h(ello|i|ey|owdy|ola)|yo|sup|good\s*(morning|afternoon|evening|night)|what'?s\s*up|gm|thanks|thank\s*you|ok(ay)?|bye|see\s*ya|cool|nice|wow|lol|haha)\b/;
-  return greetings.test(t);
+  const t = text.trim().toLowerCase().replace(/[!?.,:;]+$/g, "");
+  // Only match messages that are PURELY a greeting — no follow-up request
+  const standaloneGreetings = /^(h(ello|i|ey|owdy|ola)|yo|sup|good\s*(morning|afternoon|evening|night)|what'?s\s*up|gm|thanks|thank\s*you|ok(ay)?|bye|see\s*ya|cool|nice|wow|lol|haha)$/;
+  return standaloneGreetings.test(t);
 }
 
 async function prepareLoop(
@@ -79,7 +79,7 @@ async function prepareLoop(
   // Assemble context (uses static import — no dynamic import overhead)
   const ctx = options.systemPromptOverride
     ? { system: options.systemPromptOverride, messages: loadSession(sessionId, 50) }
-    : assembleContext(sessionId, 50, fullMemories);
+    : await assembleContext(sessionId, 50, fullMemories);
 
   const messages = compactMessages(ctx.messages, llm.contextWindow);
 

package/src/agent/prompts.ts

@@ -42,13 +42,39 @@ const DEFAULT_PERSONALITY = `You are Zubo, a personal AI agent. You are friendly
 - Use cron_create for recurring tasks. Natural language works: "every weekday at 9am", "every monday at noon".
 - Use reminder_set for one-time reminders: "in 30 minutes", "in 2 hours".
 - When the user says "remind me", "ping me", "follow up" — create a reminder.
+- Use follow_ups to schedule check-ins: "follow up about the dentist tomorrow", "check in about the project in 3 days".
+
+## Todos & notes
+
+- Use the todos tool when the user asks to track tasks, create to-do lists, or manage action items. "Add buy groceries to my list" → todos with action "add".
+- Use the notes tool to save, search, and organize information. "Save this recipe" → notes with action "save". "Find my notes about React" → notes with action "search".
+- When the user mentions something they need to do, proactively offer to add it as a todo.
+
+## Preferences
+
+- Use the preferences tool to store and recall user preferences. When the user says "I prefer morning meetings" or "I always use TypeScript" — save it immediately with the preferences tool (action "set").
+- Before making choices on behalf of the user, check their preferences first (action "get" or "list").
+- Preferences persist across sessions and channels. They're different from memory — preferences are structured key-value pairs about how the user likes things done.
+
+## Conversation topics
+
+- Use the topics tool when the user wants to organize conversations. "Let's start a new thread about the trip" → topics with action "create" then "switch".
+- When switching topics, the conversation context changes. Each topic maintains its own conversation history.
+- List topics to show the user what threads they have going.
 
 ## Delegation
 
 - Create specialized sub-agents with manage_agents for recurring task types (research, code review, data analysis).
 - Delegate tasks using the delegate tool. Sub-agents share your memory but have scoped tools.
+- When creating agents, use "*" in the tools list to give them access to all available tools dynamically (including skills and MCP tools installed later).
 - Keep the main conversation lightweight. Offload complex, self-contained tasks.
 
+## MCP (external tool servers)
+
+- You may have tools from MCP (Model Context Protocol) servers. These are prefixed with the server name, e.g. "servername__toolname".
+- MCP tools work like any other tool — just call them. Check your available tools list for MCP-provided capabilities.
+- If an MCP tool fails, it may mean the server is disconnected. Let the user know.
+
 ## Connecting services (integrations)
 
 Service integrations and LLM providers are COMPLETELY SEPARATE concepts. Never confuse them:
@@ -115,7 +141,7 @@ function loadPersonality(): string {
   return DEFAULT_PERSONALITY;
 }
 
-export function buildSystemPrompt(memories: string = ""): string {
+export async function buildSystemPrompt(memories: string = ""): Promise<string> {
   const now = new Date().toISOString();
   const personality = loadPersonality();
 
@@ -123,6 +149,28 @@ export function buildSystemPrompt(memories: string = ""): string {
 
 Current time: ${now}`;
 
+  // Inject user preferences
+  try {
+    const { loadPreferencesContext } = await import("../tools/builtin/preferences");
+    const prefsContext = await loadPreferencesContext();
+    if (prefsContext) {
+      prompt += `\n\n${prefsContext}`;
+    }
+  } catch {
+    // preferences module may not be available yet
+  }
+
+  // Inject active topic
+  try {
+    const { getActiveTopic } = await import("../tools/builtin/topics");
+    const topic = getActiveTopic();
+    if (topic) {
+      prompt += `\n\nActive conversation topic: "${topic}"`;
+    }
+  } catch {
+    // topics module may not be available yet
+  }
+
   if (memories) {
     prompt += `\n\n## Relevant memories
 <memory-data>

package/src/agent/workflow-executor.ts

@@ -3,6 +3,7 @@ import type { WorkflowDefinition, WorkflowStep } from "./workflow";
 import { getWorkflowDefinition } from "./workflow";
 import { delegateToAgent } from "./delegate";
 import { agentLoop } from "./loop";
+import { getAllToolDefs } from "../tools/registry";
 import { getDb } from "../db/connection";
 import { logger } from "../util/logger";
 
@@ -158,7 +159,10 @@ async function executeStep(
   try {
     if (step.agent === "main") {
       const sessionId = `workflow:${executionId}:${step.name}`;
-      const result = await agentLoop(llm, sessionId, task, { maxRounds: 8 });
+      const result = await agentLoop(llm, sessionId, task, {
+        maxRounds: 8,
+        allowedTools: getAllToolDefs().map(t => t.name),
+      });
       output = result.reply;
     } else {
       output = await delegateToAgent(llm, step.agent, task);