Agent Skill

Seed CLI provides an Agent Skill that teaches AI coding agents how to work with the framework — generating commands, configuring builds, using seed context modules, and following current best practices.

Once installed, your agent can help you scaffold projects, write typed commands, set up extensions and plugins, configure builds, and write tests — without needing to look up the docs manually.

Install

npx skills add SeedCLI/seedcli-skills

This works with any agent that supports the Agent Skills specification.

Supported Agents

What the Skill Covers

The skill provides your agent with accurate, source-verified knowledge about:

• Scaffolding — Project generation with create-seedcli, template types (Full, Minimal, Plugin), and project structure
• Commands — Defining commands with typed args, flags, subcommands, and the run function
• Builder API — The build() fluent API including .src(), .help(), .version(), .plugin(), .debug(), and .create()
• Seed context modules — Using print, prompt, filesystem, system, http, template, config, packageManager, strings, semver, patching, and ui from the command context
• Extensions and plugins — defineExtension(), definePlugin(), lifecycle hooks, and type augmentation via SeedExtensions
• Build and compile — seed build, seed build --compile, target platforms, --outdir, --outfile, and seed.config.ts options
• Testing — createTestCli(), mockSeed(), prompt/config/system mocking with @seedcli/testing
• Type safety — Inferred arg/flag types, typed config with defineConfig(), and typed prompt returns

How It Works

The skill uses progressive disclosure to minimize context usage:

1. Activation — Your agent reads the skill description and activates when you ask about Seed CLI
2. Core instructions — The main skill file provides quick-reference patterns for commands, builder setup, and common workflows
3. Deep references — Detailed reference files for specific modules, build configuration, and testing patterns are loaded only when needed

This means your agent gets the right level of detail for each question without wasting context on irrelevant information.

Example Prompts

Once the skill is installed, try asking your agent:

Generate a Seed CLI command with typed string args and boolean flags

Create a plugin for Seed CLI with an extension that adds a "deploy" property to the seed context

Set up seed build --compile for macOS and Windows targets

Refactor this command to use seed context modules instead of direct imports

Write tests for this Seed command using @seedcli/testing

Links

• Skill Repository
• skills.sh