a set of guidelines to build with Botpress's Agent Development Kit (ADK) - use these whenever you're tasked with building a feature using the ADK
npx skills add https://github.com/majiayu000/claude-skill-registry --skill adk
Use this skill when you've got questions about the Botpress Agent Development Kit (ADK) - like when you're building a feature that involves tables, actions, tools, workflows, conversations, files, knowledge bases, triggers, or Zai.
The Botpress ADK is a convention-based TypeScript framework where file structure maps directly to bot behavior. Place files in the correct directories, and they automatically become available as bot capabilities.
The ADK provides primitives for:
All primitives must be placed in src/ directory:
/ # Project root
├── src/
│ ├── actions/ # Strongly-typed functions → auto-registered
│ ├── tools/ # AI-callable tools → available via execute()
│ ├── workflows/ # Long-running processes → resumable/scheduled
│ ├── conversations/ # Message handlers → routes by channel
│ ├── tables/ # Database schemas → auto-created with search
│ ├── triggers/ # Event handlers → subscribe to events
│ ├── knowledge/ # Knowledge bases → RAG with semantic search
│ └── utils/ # Shared helpers (not auto-registered)
└── agent.config.ts # Bot configuration (includes integrations)
> Note: dependencies.json was removed in ADK 1.9+. All configuration including integrations now lives in agent.config.ts.
> Critical: Files outside src/ are not discovered. Location = behavior.
Activate this skill when users ask ADK-related questions like:
ADK questions fall into two categories: CLI queries and documentation lookups.
For integration discovery and CLI queries, use the Bash tool to run commands directly:
Integration Discovery:
# Search for integrations
adk search <query>
# List all available integrations
adk list --available
# Get detailed integration info (actions, channels, events)
adk info <integration-name>
# Check installed integrations (must be in ADK project)
adk list
Project Info:
# Check CLI version
adk --version
# Show project status
adk
# Get help
adk --help
When to use CLI commands:
Response pattern:
adk commandadk add [email protected] to install")For documentation, patterns, and how-to questions, search and reference the documentation files directly:
When to use documentation:
How to answer documentation questions:
pattern: **/references/*.md
pattern: <keyword from user question>
path: <path to references directory from step 1>
output_mode: files_with_matches
Documentation should be located in ./references/ directory relative to this skill. When answering questions, search for these topics:
Quick reference for accessing ADK runtime services:
// Always import from @botpress/runtime
import {
Action,
Autonomous,
Workflow,
Conversation,
z,
actions,
adk,
user,
bot,
conversation,
context,
} from "@botpress/runtime";
// Bot state (defined in agent.config.ts)
bot.state.maintenanceMode = true;
bot.state.lastDeployedAt = new Date().toISOString();
// User state (defined in agent.config.ts)
user.state.preferredLanguage = "en";
user.state.onboardingComplete = true;
// User tags
user.tags.email; // Access user metadata
// Call bot actions
await actions.fetchUser({ userId: "123" });
await actions.processOrder({ orderId: "456" });
// Call integration actions
await actions.slack.sendMessage({ channel: "...", text: "..." });
await actions.linear.issueList({ teamId: "..." });
// Convert action to tool
tools: [fetchUser.asTool()];
// Get runtime services
const client = context.get("client"); // Botpress client
const cognitive = context.get("cognitive"); // AI model client
const citations = context.get("citations"); // Citation manager
myAction.ts, searchDocs.ts (camelCase)Users.ts, Orders.ts (PascalCase)chat.ts, slack.ts (lowercase)When answering questions, always verify these patterns against the documentation:
# All package managers are supported
bun install # Recommended (fastest)
npm install # Works fine
yarn install # Works fine
pnpm install # Works fine
# ADK auto-detects based on lock files
# - bun.lockb → uses bun
# - package-lock.json → uses npm
# - yarn.lock → uses yarn
# - pnpm-lock.yaml → uses pnpm
// ✅ CORRECT - Always from @botpress/runtime
import { Action, Autonomous, Workflow, z } from "@botpress/runtime";
// ❌ WRONG - Never from zod or @botpress/sdk
import { z } from "zod"; // ❌ Wrong
import { Action } from "@botpress/sdk"; // ❌ Wrong
// ✅ Both patterns work - export const is recommended
export const myAction = new Action({ ... }); // Recommended
export default new Action({ ... }); // Also valid
// Why export const?
// - Enables direct imports: import { myAction } from "./actions/myAction"
// - Can pass to execute(): tools: [myAction.asTool()]
// ✅ CORRECT - Handler receives { input, client }
export const fetchUser = new Action({
name: "fetchUser",
async handler({ input, client }) { // ✅ Destructure from props
const { userId } = input; // ✅ Then destructure fields
return { name: userId };
}
});
// ❌ WRONG - Cannot destructure input fields directly
handler({ userId }) { // ❌ Wrong - must be { input }
return { name: userId };
}
// ✅ CORRECT - Tools CAN destructure directly
export const myTool = new Autonomous.Tool({
handler: async ({ query, maxResults }) => {
// ✅ Direct destructuring OK
return search(query, maxResults);
},
});
// ✅ CORRECT - Use conversation.send() method
await conversation.send({
type: "text",
payload: { text: "Hello!" }
});
// ❌ WRONG - Never use client.createMessage() directly
await client.createMessage({ ... }); // ❌ Wrong
When answering ADK questions, follow this structure:
Example Response Structure:
Actions are strongly-typed functions that can be called from anywhere in your bot.
**Example:**
[code example]
**Common Pitfall:** Remember to destructure `input` first (see troubleshooting section)
**Related:** You can convert Actions to Tools using `.asTool()` - see the "When to Use What" decision tree.
**Next Steps:** Create your action in `src/actions/myAction.ts` and it will be auto-registered.
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.
Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.
Replace with description of the skill and when Claude should use it.
Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies
This skill should be used when the user wants to "create a skill", "add a skill to plugin", "write a new skill", "improve skill description", "organize skill content", or needs guidance on skill structure, progressive disclosure, or skill development best practices for Claude Code plugins.
Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.
Use when creating new skills, editing existing skills, or verifying skills work before deployment
Take majiayu000/adk from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.
The instructions reference npm.
Without those the skill loads but fails at the first command.