Guide · Agent skill
How to use the Dify Workflow CLI Wrapper agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The cli-anything-dify-workflow skill teaches an AI agent to run a command-line tool for Dify workflow files. Dify workflows are visual automation pipelines for AI apps, stored as YAML or JSON. This guide explains what the skill lets the agent do, how to install it, how to ask for it in plain language, and what to check before letting it edit your files.
Dify Workflow CLI WrapperHKUDS/CLI-Anything · Apache-2.0 Download the skill Quick start
- 1Install Python packages. Install the upstream dify-ai-workflow-tools package from GitHub with pip, then install the CLI-Anything harness from the HKUDS/CLI-Anything repository.
- 2Add the skill folder. Download the skill ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
- 3Check the available commands. Run cli-anything-dify-workflow guide and list-node-types to see what the installed version supports.
- 4Create or open a workflow file. Ask the agent to create a starter file with the create command, or point it at an existing Dify YAML or JSON file.
- 5Inspect and validate. Have the agent run inspect and validate on the file, using JSON output where available, to find problems before editing.
- 6Make the edit. Describe the change, such as adding a node or setting the model, and review the resulting file changes.
- 7Review with diff. Use diff or your version control to confirm the edit before you keep it.
What an agent skill is and what this one does
An agent skill is a folder containing a SKILL.md file that tells an AI agent when and how to use a tool. This skill wraps the Dify workflow DSL CLI through a CLI-Anything harness. Once installed, the agent can call commands such as creating a starter workflow from a template, listing node types, inspecting and validating a workflow file, adding or removing nodes and connections, setting model or prompt settings, and exporting or diffing files.
All operations are local edits to Dify YAML or JSON files. The harness itself is a thin wrapper; the real workflow logic comes from the upstream open-source Dify workflow CLI that the skill depends on.
Install the skill and its dependencies
You need Python with pip and Git access, because two packages are installed directly from GitHub. Install the upstream Dify workflow CLI first, then the CLI-Anything harness.
To install the skill files, download the ZIP from this page and unzip it into ~/.claude/skills/ for your user account, or into a project's .claude/skills/ folder for one project. Alternatively, in the Claude apps go to Settings → Capabilities → Skills and upload the skill there.
- Python with pip must be available.
- Install the upstream package from the Akabane71/dify-workflow-cli repository on GitHub.
- Install the CLI-Anything harness from the HKUDS/CLI-Anything repository on GitHub.
- Git access is needed so pip can fetch both packages.
python -m pip install "dify-ai-workflow-tools @ git+https://github.com/Akabane71/dify-workflow-cli.git@main"
pip install git+https://github.com/HKUDS/CLI-Anything.git#subdirectory=dify-workflow/agent-harnessAsk the agent for workflow tasks in plain language
You do not need to type commands yourself. Describe the Dify workflow task and the agent will choose the matching command. Name the file you want changed, since the edit commands act on workflow files directly.
Good requests name the file, the action and the detail. For example: create a new workflow file from the llm template, validate an existing YAML file, or add a code node titled Process to a named workflow.
- "Create workflow.yaml as a workflow-mode starter from the llm template."
- "Inspect and validate workflow.yaml and summarise any problems."
- "Add a code node titled Process to workflow.yaml."
- "Set the model in app.yaml to openai gpt-4o."
- "Compare my current workflow with the version in git and show the diff."
Commands the agent can run
The skill documents these command groups. The agent prefers the -j or --json-output flag on upstream commands when it is available, which makes results easier for it to read.
- guide and list-node-types to see what is available
- create to make a starter file from a template
- inspect, validate and checklist to read and check a file
- edit with add-node, remove-node, update-node, add-edge, remove-edge and set-title
- config with set-model, set-prompt, add-variable, set-opening, add-question, add-tool and remove-tool
- export, import, diff and layout
cli-anything-dify-workflow guide
cli-anything-dify-workflow list-node-types
cli-anything-dify-workflow create -o workflow.yaml --mode workflow --template llm
cli-anything-dify-workflow inspect workflow.yaml -j
cli-anything-dify-workflow validate workflow.yaml -jSafety notes: scripts, network and file changes
This skill runs scripts and uses the network during installation, since it fetches packages from GitHub. Its commands edit workflow files in place, so review what the agent intends to change before it runs an edit command, and keep your files under version control so you can use diff and revert.
If dify-workflow is not on PATH but the dify_workflow Python package is installed, the wrapper falls back to running python -m dify_workflow.cli. This is documented behaviour and does not need extra setup.
- Back up or commit workflow files before letting the agent edit them.
- Use diff to review changes after an edit.
- Check the upstream repositories' code if you need to confirm what the packages do.
Tips for getting reliable results
Start with inspect and validate before making changes so the agent knows the file's current state. Ask for JSON output where it is supported, and keep one workflow per file so edits stay easy to review.
- Run guide first to see the commands the installed version offers.
- Use list-node-types to confirm the exact node type names before adding nodes.
- Use export and import when you need to move a workflow between formats.
What people use it for
- Creating a starter workflow from the llm template and filling in its settings.
- Validating and summarising problems in a Dify YAML file kept in a git repository.
- Adding, removing or renaming nodes and connections across a workflow without manual editing.
- Setting model provider, model name or prompt text in an app's YAML file.
- Comparing two versions of a workflow file with diff before a code review.
Frequently asked questions
How do I install the cli-anything-dify-workflow skill?
Download the skill ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps. You also need to install the two Python packages from GitHub listed in the installation section.
What do I need to run this skill?
You need Python with pip, Git access so pip can install packages from GitHub, the upstream dify-ai-workflow-tools package from the Akabane71/dify-workflow-cli repository, and the CLI-Anything harness from the HKUDS/CLI-Anything repository.
Is the Dify Workflow CLI Wrapper skill free?
This page does not state a licence or price for the skill itself. The source of the underlying code, HKUDS/CLI-Anything, is listed here under Apache-2.0, so check that repository and the upstream packages for their terms.
Does this skill work without Dify?
No. The commands only make sense for Dify's DSL format, which stores workflows as YAML or JSON files. If you do not use Dify, this skill will not help.
Can the agent change my workflow files without asking?
Edit commands such as add-node, remove-node and set-model change workflow files directly. Keep your files under version control and review the diff after each edit.
How does this compare with editing Dify workflows in the visual editor?
This skill is a command-line wrapper with no graphical interface. It suits repetitive edits and version-controlled YAML, while the visual Dify editor suits people who prefer to build workflows by clicking.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
This skill suits developers and teams who keep Dify workflow YAML or JSON in version control and want an agent to run validation, diffs and repetitive edits from the command line. It is a poor fit for anyone who wants a no-code interface, does not use Dify, or cannot review file changes before the agent makes them.
Download the skill