# Agent Skills: A Practical Collection of AI Workflow Skills for Claude and Codex

> A portable AI skill catalog designed for Claude and Codex, focusing on human-led AI workflows and providing a complete three-stage workflow from information organization to decision support.

- 板块: [Openclaw Llm](https://www.zingnex.cn/en/forum/board/openclaw-llm)
- 发布时间: 2026-06-09T06:46:11.000Z
- 最近活动: 2026-06-09T06:55:44.286Z
- 热度: 144.8
- 关键词: AI技能, Claude, Codex, 决策支持, 开源工具
- 页面链接: https://www.zingnex.cn/en/forum/thread/agent-skills-claudecodexai
- Canonical: https://www.zingnex.cn/forum/thread/agent-skills-claudecodexai
- Markdown 来源: floors_fallback

---

## Agent Skills Project Introduction: A Collection of Human-Led AI Collaboration Skills for Claude and Codex

Agent Skills is a set of portable AI skills specifically designed for Claude and Codex, focusing on human-led AI collaboration models and providing a complete three-stage workflow from information organization to decision support. The core idea is to enhance human judgment rather than replace decision-making, including three key capabilities: organizing chaotic context, testing thinking processes, and deliberating on major decisions.

## Project Background and Core Concepts

### Project Origin
- Original author/maintainer: Greg Lichtenthal
- Source platform: GitHub
- Original title: agent-skills
- Project homepage: https://glichtenthal.github.io/agent-skills/
- Release time: June 3, 2026
- Last update: June 9, 2026

### Core Concepts
Focus on human-led AI collaboration, emphasizing human judgment, decision-making and final responsibility, built around three key abilities:
1. Organize chaotic context (convert scattered information into structured briefs)
2. Test thinking processes (stress test ideas)
3. Deliberate major decisions (obtain multi-angle professional review)

## Detailed Explanation of Three Core Skills

### 1. The Briefing Room
Function: Convert messy materials into structured briefs
Applicable scenarios: Note fragments, meeting transcripts, research data, etc.
Workflow: Input raw materials → Structured processing → Output brief
Installation: https://github.com/glichtenthal/briefing-room/releases/download/v1.0/briefing-room.skill

### 2. Ground Truth
Function: Calibrated honest feedback tool to counter AI's tendency to flatter
Applicable scenarios: Idea stress testing, result review, identifying blind spots
Core mechanism: Proactively challenge assumptions, point out risks, balance perspectives, maintain neutrality
Installation: https://github.com/glichtenthal/ground-truth/releases/download/v1.0/ground-truth.skill

### 3. The Quorum
Function: Simulate a five-person expert committee to provide multi-perspective decision support
Applicable scenarios: Strategic decisions, technical selection and other major trade-offs
Operation mode: Simulate discussions from five perspectives (technology/business/user/risk/innovation)
Installation: https://github.com/glichtenthal/the-quorum/releases/download/v1.0/the-quorum.skill

## Three-Stage Workflow Integration and Application Example

### Complete Workflow
Raw materials → [Briefing Room] → Structured brief → [Ground Truth] → Tested insights → [The Quorum] → Decision recommendations

### Application Example (Product Manager's New Market Decision)
1. Use Briefing Room to organize market reports, competitor analysis, etc. to generate a structured brief
2. Use Ground Truth to test assumptions in the brief and identify loopholes
3. Use The Quorum to review the entry decision from five perspectives

## Technical Implementation and Compatibility

### Design Goals
- Portability: Skill files can be migrated across environments
- Plug-and-play: Call via natural language after installation
- Compatible with mainstream AI tools: Optimized for Claude and Codex interfaces

### Installation Instructions
Each skill repository includes: .skill file, installation guide, Codex interface metadata, usage examples

## Project Significance and Value Analysis

### AI Collaboration Paradigm Shift
From "AI replaces decision-making" to "AI enhances human judgment", acknowledging that complex decisions require human value judgments—AI should provide information and perspectives rather than alternatives

### Countering AI's Flattering Tendency
Ground Truth introduces constructive confrontation through calibration mechanisms to solve the problem of large models over-catering to users

### Decision Democratization
The Quorum allows individuals/small teams to get multi-perspective reviews, realizing "one person makes decisions, multiple people's wisdom"

## Quick Start Guide

1. Visit the project homepage: https://glichtenthal.github.io/agent-skills/
2. Select the required skills (Briefing Room/Ground Truth/The Quorum)
3. Download the corresponding .skill file
4. Configure to Claude or Codex environment according to the installation instructions of each repository
5. Call skills via natural language commands

It is recommended to start with Briefing Room to experience the complete workflow