Vorlaxen Agent Skills

Decision boundaries for AI coding agents — what they may decide, what they must not, and when to stop and ask.

Works with Cursor, Claude Code, Copilot, Windsurf, Codex, Gemini CLI, and any tool that reads custom instructions.

Maintained by Vorlaxen Labs as both our internal agent config and an open community standard.

What is this?

This repository is two things in one:

Not Vorlaxen-specific? Copy AGENTS.md and ignore the library folder.

Using Vorlaxen packages? Add the matching library skill on top of the global standards.

The @vorlaxen-labs/agent-skills CLI installs the right files into your project for your platform — interactively or in CI. Full CLI reference lives in docs/.

Core idea

Constrain decisions, not implementations.

These rules tell an agent what it may decide — not how to write code. Stack, patterns, and folder structure come from your project.

Four hard boundaries:

1. Never hallucinate — if it's not in the codebase, it doesn't exist
2. Match risk to action — ask for API contracts, schema, security; proceed when the pattern is already in the code
3. Scope control — no silent refactors or infrastructure changes
4. Follow the project — your conventions beat agent preference

Full rules: AGENTS.md

What's included

Engineering standards

Universal modules for every project — global rules, optional web frontend/backend, and audit modules such as kill-ai-slop. See Included skills.

Vorlaxen library skills

Optional skills for @vorlaxen-labs/bar-js, @vorlaxen-labs/huk-js, and @vorlaxen-labs/kargomucuz-sdk. Each ships with a SKILL.md entry point and a reference/ folder with full API documentation verified against source.

Get started

npx @vorlaxen-labs/agent-skills init

Contributing

Contributions welcome — see CONTRIBUTING.md. Skill authoring spec: SKILL-SPEC.md.

License

MIT — see LICENSE. Copyright (c) 2026 Vorlaxen Labs.