Recipe Skills

An open-source knowledge library for agent-assisted Workato recipe development. A Workato Labs project.

What are Recipe Skills?

Recipe Skills teach AI agents how to generate valid Workato recipe JSON. Each skill encapsulates the connector-specific knowledge an agent needs -- which actions exist, when to use each one, how datapills work for that connector, and what gotchas to avoid. Without skills, agents hallucinate action names, produce malformed schemas, and miss critical connector rules.

A skill IS:

• A knowledge bundle that teaches agents how to write recipes for a specific connector
• Documentation, validated templates, and patterns an agent can reference during generation
• Audited lint rules defining every valid action and trigger name for the connector

A skill is NOT:

• A code library you import at runtime
• Pre-packaged operations or a connector SDK
• A replacement for the Workato platform

Available Skills

Base Skill

Skill	Description
workato-recipes	Core Workato recipe fundamentals -- triggers, control flow, datapills, formulas, recipe JSON structure

Connector Skills

Skill	Connector	Native Actions	Triggers
asana-recipes	Asana	16	2
gmail-recipes	Gmail	3	1
jira-recipes	Jira	17	11
salesforce-recipes	Salesforce	32	15
slack-recipes	Slack (native + Workbot)	9	2
stripe-recipes	Stripe	7	4

Action and trigger counts reflect what is audited in each skill's lint-rules.json. All connectors also support __adhoc_http_action for API operations beyond the native action set.

Quick Start

With Claude Code

This repo includes slash commands for each skill. From this repo directory:

/workato-recipes    # Load recipe fundamentals
/stripe-recipes     # Generate Stripe recipes
/salesforce-recipes # Generate Salesforce recipes
/slack-recipes      # Generate Slack/Workbot recipes
/gmail-recipes      # Generate Gmail recipes
/jira-recipes       # Generate Jira recipes
/asana-recipes      # Generate Asana recipes

Then ask the agent to generate a recipe:

• "Create an API endpoint recipe that accepts customer data"
• "Create a recipe to search for a Stripe customer by email and create a PaymentIntent"
• "Create a recipe to upsert a Salesforce Contact by external ID"
• "Create a slash command that collects user input via dialog and posts to Slack"
• "Create a recipe that searches Jira issues by JQL and returns sprint status"
• "Create a recipe that creates an Asana task with subtasks"

The agent uses the skill's documentation, templates, and lint rules to generate valid recipe JSON.

With Other AI Agents (Codex, Cursor, Windsurf, etc.)

Each skill directory contains a SKILL.md file following the open agent skills standard. Any SKILL.md-aware tool can discover and load these skills automatically.

To load skills manually, point your agent to the entry points:

1. Load the base skill first:

   Read skills/workato-recipes/SKILL.md
   

2. Then load a connector skill:

   Read skills/<connector>-recipes/SKILL.md
   

Each skill.yaml defines an entry_point and extends: workato-recipes to declare the base dependency. Agents or tooling can parse this metadata to automate skill loading.

Deploy

Generated recipes are standard Workato recipe JSON. Push them to your workspace:

wk push

Skill Structure

Each connector skill follows this structure:

skills/<connector>-recipes/
├── skill.yaml              # Metadata and manifest
├── SKILL.md   # Main knowledge document
├── lint-rules.json         # Valid action/trigger names (source of truth)
├── validation-checklist.md # Connector-specific validation checks
├── templates/              # Validated recipe templates
│   └── *.json
└── patterns/               # Pattern documentation
    └── *.md

The base skill (workato-recipes) additionally includes fundamentals/, triggers/, and control-flow/ subdirectories covering core Workato concepts that all connector skills inherit.

How It Works

Skills are structured around a key principle: lint rules own the "what," instructions own the "how."

• lint-rules.json defines every valid action and trigger name for the connector, audited against the actual Workato connector. This is the single source of truth that prevents agents from hallucinating action names.

• SKILL.md teaches agents when and how to use each action through decision-logic guidance (e.g., "use search_sobjects for exact match, search_sobjects_soql for complex SOQL queries"). It references lint-rules.json for valid names rather than duplicating them.

• templates/ provide complete, validated recipe JSON that agents can reference and adapt. These are real recipes that have been pushed to Workato, not placeholder scaffolds.

• validation-checklist.md gives agents a pre-push verification checklist covering connector-specific rules that lint alone cannot catch (datapill path formats, EIS requirements, query syntax).

Documentation

Skill Instructions

• workato-recipes -- Recipe fundamentals and base patterns
• asana-recipes -- Asana task and project management
• gmail-recipes -- Gmail email operations
• jira-recipes -- Jira issue tracking and JQL
• salesforce-recipes -- Salesforce CRM operations
• slack-recipes -- Slack and Workbot integration
• stripe-recipes -- Stripe payment processing

Project

• Roadmap -- Project direction, connector prioritization, how to get involved
• Contributing Guide -- How to add or improve skills
• wk CLI + Recipe Lint Guide -- Linter setup, tiers, rule reference

Contributing

We welcome contributions. See CONTRIBUTING.md for the full guide.

• Add a new connector skill -- build a skill for a connector you use
• Improve existing skills -- fix bugs, add patterns, add templates
• Extend the base skill -- improve core recipe knowledge that all connector skills inherit

License

MIT License -- see LICENSE for details.

Support

• Issues: GitHub Issues