< woml />
View on GitHub

Build with AI

The WOML Skill teaches compatible AI coding agents how to turn a plain-language automation request into a valid, readable, and runnable .woml file. It gives the agent the released WOML tags, runtime bindings, services, providers, CLI commands, reliability rules, and common workflow patterns.

Instead of explaining WOML syntax in every prompt, describe the outcome you want. The agent can choose an appropriate trigger and control flow, create the workflow, identify required secrets, validate the file, and tell you exactly how to run it.

Install the skill

Install the skill inside a project when you want everyone working in that repository to use the same WOML guidance.

Claude Code

Run these commands from your project directory:

Terminal
mkdir -p .claude/skills/woml
curl -fsSL https://github.com/dali-benothmen/woml/releases/latest/download/woml-skill.tar.gz \
  | tar -xz -C .claude/skills/woml

Codex and Agent Skills-compatible tools

If you have cloned the WOML repository, copy the complete skill directory into your project:

Terminal
mkdir -p .agents/skills
cp -R /path/to/woml/skills/woml .agents/skills/woml

Keep the complete woml directory intact. Its reference files contain the released language, service, provider, module, diagnostic, and command contracts that the agent uses while building workflows.

Restart or reload your coding agent if it does not discover the newly installed skill immediately.

Ask for a workflow

You can ask naturally:

Text
Build a WOML workflow that receives an order through a webhook, validates it,
checks inventory and fraud risk concurrently, asks for approval when the risk
is high, and returns an order decision. Save it as order-processing.woml.

Clients that support explicit skill invocation can use $woml:

Text
$woml Create a daily report workflow that loads sales data, calculates totals,
and sends the summary to Telegram.

Include any requirements that matter to your automation, such as the trigger, expected payload, external services, retry behavior, approval rules, or desired output. You do not need to choose every WOML tag yourself.

What the agent should deliver

A successful WOML Skill request should produce:

  • One or more readable .woml files.
  • Supporting JavaScript or TypeScript modules only when the workflow needs them.
  • A list of required secrets without exposing secret values.
  • The commands needed to validate and run the workflow.
  • Clear trigger instructions, such as pressing Enter or calling a webhook URL.
  • An explanation of important assumptions or runtime behavior.

When the WOML CLI is available, the agent should validate generated files with:

Terminal
woml check order-processing.woml

Validation catches unsupported tags, invalid attributes, duplicate IDs, broken references, and incompatible workflow structure before execution.

Review before running

AI-generated workflows are code. Review scripts, module imports, service calls, filesystem access, and notification recipients before running them, especially when a workflow can modify data or contact an external system.

Store credentials with woml secrets set and reference them through secrets. Never paste real tokens or passwords into a prompt or commit them inside a .woml file.

Continue learning