playbooks

🥈Silver

Playbooks is a semantic programming system for AI agents. It enables operations teams to automate workflows using natural language. Connects to Claude agents for task execution and orchestration.

81100Updated 2w ago

Intermediate30min to implementautomation

Saves ~480 min per use

Quick InstallView Source

git clone https://github.com/playbooks-ai/playbooks.git

Works with:

Claude

Overview

About This Skill

Playbooks is a semantic programming system for AI agents. It enables operations teams to automate workflows using natural language. Connects to Claude agents for task execution and orchestration.

How to Use

[{"step":"Define the workflow","action":"Identify the end-to-end process you want to automate (e.g., customer onboarding, incident response, lead qualification). Break it down into 3-5 high-level steps. Use tools like Lucidchart or Miro to map the workflow visually before converting it into a playbook.","tip":"Start with a process that is repetitive, time-consuming, and has clear success criteria. Avoid automating ambiguous or highly variable workflows initially."},{"step":"Customize the playbook template","action":"Replace the [PLACEHOLDERS] in the prompt template with your specific steps, tools, edge cases, and success criteria. For example, if automating incident response, your steps might include 'Triage the alert,' 'Assign to the on-call engineer,' and 'Notify stakeholders.'","tip":"Use the Claude Code Interpreter to validate the playbook logic (e.g., check for missing steps or invalid tool configurations) before deployment."},{"step":"Deploy to a Claude agent","action":"Copy the customized playbook into your Claude agent's prompt and configure the necessary API integrations (e.g., Slack, Jira, AWS). Use the `tools` parameter to grant the agent access to the required APIs and permissions.","tip":"Test the playbook in a staging environment first. Use the `execute` command in Claude to run a dry-run and verify the agent's responses."},{"step":"Monitor and iterate","action":"Track the playbook's performance using the success criteria (e.g., completion time, error rates). Use tools like Datadog or Grafana to log metrics and identify bottlenecks. Update the playbook as needed based on feedback and failures.","tip":"Schedule a weekly review of the playbook's performance with your operations team. Use this time to refine edge case handling and optimize tool configurations."},{"step":"Scale across teams","action":"Once the playbook is stable, share it with other teams (e.g., sales, support) by exporting the playbook as a JSON file and importing it into their Claude agents. Use version control (e.g., Git) to manage updates and ensure consistency.","tip":"Create a shared playbook library in your organization's documentation tool (e.g., Confluence, Notion) where teams can contribute and reuse playbooks."}]

Setup & Installation

Quick Install

No install command available. Check the GitHub repository for manual installation instructions.

Alternative Install (Git Clone)

git clone https://github.com/playbooks-ai/playbooks

Requirements

Claude Code or compatible AI agent
Works with: Claude

Quick Start Guide

Install the Skill

Copy the install command above and run it in your terminal.

Open Your AI Agent

Launch Claude Code, Cursor, or your preferred AI coding agent.

Try It Out

Use the prompt template or examples below to test the skill.

Customize

Adapt the skill to your specific use case and workflow.

Usage Examples

Prompt Template

Design an AI agent playbook for [TASK/PROCESS]. The playbook should include: 1) [STEP_1], 2) [STEP_2], 3) [STEP_3], and 4) [STEP_4]. Use [TOOLS] where applicable. Ensure the playbook handles [EDGE_CASES] and includes [SUCCESS_CRITERIA]. Deploy this playbook to a Claude agent for execution.

Example Output

### AI Agent Playbook: Customer Onboarding Automation

**Objective:** Automate the customer onboarding process for SaaS subscriptions, reducing manual intervention by 70% and ensuring 95% onboarding completion within 24 hours.

**Playbook Steps:**

1. **Trigger & Validation**
- *Trigger:* New customer signup via Stripe webhook (event: `customer.subscription.created`).
- *Validation:* Verify payment status is `active` and customer email is valid. If invalid, log error and notify Slack channel `#onboarding-errors`.
- *Tools:* Stripe API, Slack API, Claude agent.

2. **Provisioning**
- *Action:* Create customer account in HubSpot CRM and assign to `onboarding` pipeline.
- *Action:* Generate and send personalized welcome email via SendGrid (template: `welcome_onboarding`).
- *Action:* Provision AWS resources (EC2 instance, RDS database) using Terraform Cloud API.
- *Tools:* HubSpot API, SendGrid API, Terraform Cloud.

3. **Configuration & Training**
- *Action:* Schedule a 30-minute Zoom meeting with the customer using Calendly API (template: `onboarding_training`).
- *Action:* Send a pre-meeting checklist to the customer (via email) with links to product documentation and a 2-minute tutorial video.
- *Action:* Log the meeting details in Notion (database: `Customer Onboarding`).
- *Tools:* Calendly API, SendGrid API, Notion API.

4. **Completion & Handoff**
- *Action:* Mark onboarding as `completed` in HubSpot CRM and update `onboarding_status` to `success`.
- *Action:* Send a post-onboarding survey to the customer (via Typeform) with a 5-star rating and feedback form.
- *Action:* Notify the sales team in Slack (`#onboarding-alerts`) with a summary of the onboarding session.
- *Tools:* HubSpot API, Typeform API, Slack API.

**Edge Cases & Handlers:**
- *Payment Failure:* Retry payment once after 2 hours. If still failed, escalate to finance team via email.
- *Provisioning Failure:* Roll back AWS resources using Terraform Cloud and notify the engineering team via PagerDuty.
- *Customer No-Show:* Reschedule the Zoom meeting automatically via Calendly and notify the customer via email.

**Success Criteria:**
- 95% of onboarding processes complete within 24 hours.
- 90% customer satisfaction score (from Typeform survey).
- Zero manual intervention required for 80% of onboarding tasks.

**Deployment:**
This playbook is deployed to a Claude agent with the following permissions:
- Access to Stripe, HubSpot, SendGrid, Terraform Cloud, Calendly, Notion, Typeform, Slack, and PagerDuty APIs.
- Permissions to execute Terraform plans and log data in Notion.
- Automated retries for transient failures (e.g., API timeouts).

**Metrics Dashboard:**
Track onboarding completion time, failure rates, and customer satisfaction scores in a shared Google Sheet (updated hourly via Zapier).