Agent Skills

Teach your AI agent to create and manage Schematify graphs

Agent Skills are instruction files that teach an AI agent how to work with Schematify. They tell the agent when to use the Schematify CLI, how to inspect a project, how to assemble graph data, and how to push the result to Schematify.

The skills live in the public schematify/skills repository.

Available skills

SkillUse it for
schematify-cliOperating the schematify CLI. The agent uses it for listing graphs, pulling documents, pushing documents, deleting documents, converting graph inputs, and checking CLI help.
schematify-graphsAuthoring and editing Scriptable Graph TypeScript files with the graph, node, channel, from, and channelPublisher APIs.
schematify-generateGenerating a graph from a codebase, project, directory, dataset, or freeform description. The agent works out what to map, sources nodes and links, then hands the graph shape to the graph-authoring workflow.
schematify-renderConfiguring property cards, reports, and charts through the graph builder’s .render(...) API.

The four skills are designed to compose. schematify-cli operates the tool, schematify-graphs authors graph scripts, schematify-generate discovers what should be in the graph, and schematify-render configures specialized node rendering.

Install

Skills install with the CLI installer. After the CLI is installed, the installer prompts:

Install AI agent skills? [Y/n]

Choose Yes (the default) to install all four skills globally into:

  • ~/.schematify/skills/ (canonical copy)
  • ~/.agents/skills/
  • ~/.claude/skills/
  • ~/.cursor/skills/
  • ~/.codex/skills/

No Node.js is required. Re-running the installer overwrites the Schematify skill directories and leaves other skills untouched.

Headless:

curl -fsSL https://<your-schematify-host>/install.sh | sh -s -- --install-skills
curl -fsSL https://<your-schematify-host>/install.sh | sh -s -- --skip-skills

You can also clone or download schematify/skills and place the skill files wherever your agent expects them.

Prerequisites

Install the Schematify CLI (the same installer can install skills). For the hosted Schematify service, the CLI has a default configuration, so setup is not required. Authenticate with:

schematify login

Run setup only for a custom VM or non-default deployment:

schematify setup

What the CLI skill does

The schematify-cli skill tells the agent to use the CLI as the source of truth for Schematify resources. It should not search your filesystem for graph state when it can ask the server through the CLI.

It also tells the agent to:

  • run schematify --help and command-specific help instead of guessing flags
  • use --json when it needs machine-readable output
  • rely on command exit codes for actions like push, delete, and convert
  • use the installed CLI binary rather than reimplementing server calls

Use it for requests like:

  • “List my Schematify graphs”
  • “Pull the production architecture graph”
  • “Push this document”
  • “Delete the old test graph”
  • “Convert this interchange JSON”

What the graphs skill does

The schematify-graphs skill is for authoring and editing Scriptable Graph files. It tells the agent how to use the builder API, preserve graph IDs, keep node IDs stable, add links with absolute node paths, define channels, and publish live channel values.

Use it for requests like:

  • “Add a Redis node to this graph script”
  • “Wire the API node status to a channel”
  • “Add live channel updates to this graph”
  • “Create a Scriptable Graph for this topology”

What the render skill does

The schematify-render skill defines the parameters, bindings, and sizing rules for property cards, reports, pie charts, bar charts, and line charts. Agents use it with schematify-graphs whenever a graph script calls .render(...).

What the generate skill does

The schematify-generate skill is for deciding what should go into a graph. It starts by deciding what kind of graph you want:

  1. Architecture diagram: logical components and how they communicate.
  2. Code dependency graph: source files and their import relationships.
  3. Custom graph: a graph based on your description.

For architecture graphs, the agent scans the project for major components, services, modules, databases, routes, and runtime dependencies. It should produce a whiteboard-style model, not a file listing.

For dependency graphs, the agent maps source files and internal imports. This is intentionally more detailed and file-oriented.

For custom graphs, the agent follows your description and asks clarifying questions if the target is vague.

Generation workflow

The current generate workflow is:

  1. Identify what should be graphed.
  2. Inspect the relevant project, folder, or information source.
  3. Assemble the nodes, hierarchy, links, and metadata.
  4. Author or update a Scriptable Graph using the graph builder API.
  5. Run the graph with the CLI to validate or publish it.

For workflows that still use interchange JSON, the agent can convert it with schematify convert. For Scriptable Graph workflows, the generated structure is represented directly in TypeScript.

Example prompts

Create a Schematify architecture graph of this project.
Generate a dependency graph for apps/client/src.
Map our deployment topology: one frontend, three backend services, Redis, PostgreSQL, and the external payment provider.
Pull the existing infrastructure graph and add the new worker service.

What’s next