A collection of plugins for agentic AI assistants, built on top of composable "skills" and instructions that make sure your agent uses them.
Create, review, customize, and manage plugins and reusable agent artifacts. Includes workflows for scaffolding plugins, writing skills against a shared standard, reviewing skill bundles and subagents, and adapting distributable plugins for your organization.
A complete software engineering workflow: brainstorming, planning, TDD, debugging, code review, and more. Built on composable skills that trigger automatically when your agent encounters relevant tasks.
A collection of skills that enhance AI agents with specialized capabilities for developing on the Moonbeam parachain: adding pallets, precompiles, XCM messaging, runtime development, testing, migrations, and more.
Note: Installation differs by platform. Claude Code and Cursor have built-in plugin marketplaces. Codex and OpenCode require manual setup.
In Claude Code, register the marketplace first:
/plugin marketplace add Moonsong-Labs/knowledge-work-pluginsThen install the plugin from this marketplace:
/plugin install core-engineering@moonsong-labsCursor automatically detects plugins installed by Claude Code. Install via Claude Code first, then restart Cursor.
Tell Codex:
Fetch and follow instructions from https://raw.githubusercontent.com/Moonsong-Labs/knowledge-work-plugins/refs/heads/main/core-engineering/.codex/INSTALL.md
Detailed docs: core-engineering/docs/README.codex.md
Tell OpenCode:
Fetch and follow instructions from https://raw.githubusercontent.com/Moonsong-Labs/knowledge-work-plugins/refs/heads/main/core-engineering/.opencode/INSTALL.md
Detailed docs: core-engineering/docs/README.opencode.md
Start a new session in your chosen platform and ask for something that should trigger a skill (for example, "help me plan this feature" or "let's debug this issue"). The agent should automatically invoke the relevant skill.
- Test-Driven Development - Write tests first, always
- Systematic over ad-hoc - Process over guessing
- Complexity reduction - Simplicity as primary goal
- Evidence over claims - Verify before declaring success
To run the root just check and just test recipes locally, install these CLIs and make sure they are on your PATH:
justbun(providesbunxformarkdownlint-cli2andshellcheck)go(used viago runforshfmtandactionlint)yamllinttyposskills-refclaude(Claude Code CLI)jqnodeandnpmpython3git- GNU
timeout(coreutils; on macOS you may need to install it separately sotimeoutis available inPATH)
You do not need to install markdownlint-cli2, shellcheck, shfmt, or actionlint separately because the root justfile runs them through bunx and go run.
The Claude-based test suites also expect the local plugin to be enabled. This can be most easily be enabled via the --plugin-dir command line option.
The writing-skills skill in the plugin-management plugin guides you through creating and testing new skills.
-
Fork and clone the repository
-
Start Claude Code with the plugin-management plugin loaded:
claude --plugin-dir ./plugin-management
-
Ask your agent to create a new skill:
I want to add a new skill to [plugin name] for [describe what it does]The
writing-skillsskill will activate and guide you through the TDD process: write failing tests (baseline behavior), write the skill, verify it works, and refactor. -
Submit a PR with your new skill.
Plugins are self-contained directories at the root of this repository. The plugin-management plugin provides a guided workflow for creating new plugins.
-
Fork and clone the repository
-
Start Claude Code with the plugin-management plugin loaded:
claude --plugin-dir ./plugin-management
-
Ask your agent to create a new plugin:
I want to create a new plugin for [describe what it does]The
create-pluginskill will activate and walk you through a 5-phase process: discovery, component planning, designing your first component, generating all files, and validation. -
Submit a PR with your new plugin.
Skills update automatically when you update the plugin:
/plugin update core-engineeringInspired by: