Workato Developer Skills

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

What are Workato Developer Skills?

Workato Developer 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

Connector Skills

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

Recipe Skills install as a plugin in your coding agent. Install it once, then ask for recipes in any project — you don't clone this repo or work inside it. The plugin bundles the base workato-recipes skill plus every connector skill; your agent loads the base and the relevant connector automatically when you ask for a recipe.

Install the plugin

Claude Code

/plugin marketplace add workato-devs/recipe-skills
/plugin install recipe-skills@workato-labs

Codex

codex plugin marketplace add https://github.com/workato-devs/recipe-skills

Then open /plugins, install recipe-skills, and restart your session.

Cursor — coming soon (pending publication to the Cursor plugin marketplace).

Use it

Ask for a recipe in plain language:

• "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"

Tip: name your connection when it matters (e.g. "…using my Gmail connection named 'My Gmail account'") — the agent can't guess connection names, and importing a recipe fails if the named connection doesn't exist in your workspace.

Working in this repo (contributors)

If you're improving the skills themselves, clone the repo and open it in any AGENTS.md-aware agent (Codex, Cursor, and others) — the root AGENTS.md routes a request to the right skill. Each skill follows the open agent skills standard (name/description frontmatter; skill.yaml with extends: workato-recipes). See CONTRIBUTING.md.

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