LabKit

English | 中文

License PRs Welcome Last Updated

Reusable agent skills for publishing, everyday work, and configurable Astra behavior.

New: Welcome to AGI packages five official GPT-6 Astra prompt modules with individual switches and ordinary-task semantic routing. First-use guidance presents module/hook status and asks about module changes, an optional Skills conflict check and routing preferences. Initialization guide. Host guide. 中文介绍

Install

git clone https://github.com/haorantang97/LabKit.git
cd LabKit

LabKit exposes four top-level skills. skill-skill is one complete publishing toolkit; its pipeline modules live under skills/skill-skill/modules/ and are not separate top-level skills. Astra's behavior modules are also internal to one skill.

Each top-level skill has a SKILL.md entrypoint. Install the complete folder, including any scripts and references. For Welcome to AGI in a local Codex project:

python3 skills/welcome-to-agi/scripts/install.py --host codex --surface desktop --project /path/to/project --apply

Use --user instead of --project for user-wide installation. This example initializes a persistent rule; /hooks trust is only needed if you explicitly select hook mode. Check existing installations before copying to preserve customizations. For other hosts, follow their skill installation instructions and retain referenced files; copying SKILL.md over AGENTS.md loses resources and can overwrite project rules. Welcome to AGI also provides Claude Code and Cursor rule adapters; see its host guide.

The four top-level skills are independent. Install only the ones you want.

What it does

LabKit is a small collection of four skills. skill-skill is the publishing toolkit; open-loops and reading-plan-mentor are daily-use tools. welcome-to-agi combines attributed official prompt blocks with modular routing and optional setup tools.

Skills

  • welcome-to-agi — Routes ordinary tasks through selected official guidance; includes modular prompts, host-aware initialization, and optional configuration audit

  • skill-skill — Packages a working AI rule or skill and guides it through publication as a GitHub repository

  • open-loops — Audits long conversations for unanswered, unacknowledged, or assistant-decided information points

  • reading-plan-mentor — Turns a book list into a paced, long-term reading plan with daily guidance and continuity

These two skills are also maintained in their standalone repositories: open-loops and reading-plan-mentor.

File format

skills/{name}/SKILL.md

YAML front matter (name, description, license) followed by markdown:

---
name: skill-name
description: "Use when the skill's triggering conditions match the user's request."
license: PolyForm-Noncommercial-1.0.0
---

## Quick Start
...

Contributing

See CONTRIBUTING.md.

License

PolyForm Noncommercial License 1.0.0. See LICENSE.