Wiki Overview

中文 | English

Purpose

This vault turns files placed in raw/ into durable markdown pages that an LLM can maintain over time.

The wiki is intended to accumulate synthesis, not to recompute the same understanding from scratch on every query.

Start Here

• LLM Wiki - Core pattern for maintaining a persistent synthesis layer instead of re-deriving answers from raw files.
• Knowledge Operations - Canonical ingest, query, lint, and raw-refactor loop for maintaining this vault.
• Vault Architecture - Three-layer model that explains how raw/, wiki/, index.md, and log.md fit together.

Core Topics

• Agent-First Repositories - How repositories become readable control planes for autonomous coding agents.
• Agent-Readable Design Systems - How markdown design contracts and style references constrain frontend work.
• Agent Harness Design - Durable patterns for planner-generator-evaluator loops, evaluation, and simplification.
• Codex Operating Practices - Practical operating model for Codex configuration, skills, plugins, and execution modes.

Durable Answers

• How To Make Agents Run Longer - Practical checklist derived from the current harness and repository sources.

Operations

• OpenClaw Field Notes - Maintained incident and troubleshooting hub for real OpenClaw deployment lessons.
• OpenClaw Skills Symlink Escape - Why escaped skill symlinks can look installed while the official loader still skips them.
• OpenClaw Cron Pairing And Operator Scopes - How a pairing-flavored cron failure turned out to be an operator-scope problem.
• OpenClaw Provider Endpoint DNS Failures - How to trace GPT-side DNS failures back to the active provider endpoint config.
• Raw Markdown Build-Safe Linking - How transplanted README-style raw markdown can break local builds when upstream-relative links are interpreted in the wrong repository context.
• OpenClaw Field Note Template - Shared writing template and capture threshold for turning new incidents into durable operator notes.
• OpenClaw Field Note Capture Loop - Concrete pattern for keeping reusable operator lessons from disappearing into chat history.

Source Families

Anthropic / Claude Code / Everything Claude Code

• Affaan Mustafa Claude Code Shorthand Guide
• Affaan Mustafa Claude Code Longform Guide
• Everything Claude Code GitHub Repository

Anthropic

• Anthropic Harness Design For Long-Running Application Development
• Claude Managed Agents Overview
• Scaling Managed Agents: Decoupling The Brain From The Hands
• Claude Code Auto Dream Memory Consolidation
• Claude Mythos Preview Cybersecurity Assessment
• Claude Mythos Preview System Card
• Introducing Claude Opus 4.7

OpenAI / Codex

• OpenAI Harness Engineering In An Agent-First World
• Codex Best Practices
• Codex CLI Slash Commands
• Codex Non-Interactive Mode
• Codex Agent Skills
• Codex Subagents
• Codex Plugin For Claude Code
• GitHub OpenAI Codex Repository

OpenAI / Codex / Configuration

• Codex Config Basics
• Codex Advanced Configuration
• Codex Configuration Reference
• Codex Sample Configuration

OpenAI / Codex / Skills

• Codex LLM Wiki Skill

Karpathy

• Karpathy LLM Wiki Gist
• Karpathy Claude Coding Thread
• Forrestchang Andrej Karpathy Skills Repository

Cocoon AI

• Architecture Diagram Generator

Daniel Breunig

• How Claude Code Builds a System Prompt

EveryInc

• Compound Engineering Plugin

Garry Tan

• gstack

Multica AI

• Multica

Piebald AI

• Claude Code System Prompts Repository

Shanraisshan

• Claude Code Best Practice Repository
• Claude Code Best Practice Tips Compendium
• Codex CLI Best Practice

Ralph

• Ralph Wiggum Loop Technique
• Ralph Step-By-Step Guide
• Ralph GitHub Repository

VoltAgent

• VoltAgent Awesome DESIGN.md

Suggested Paths

• Harness path: Agent Harness Design -> Anthropic Harness Design For Long-Running Application Development -> How To Make Agents Run Longer
• Managed Agents path: Claude Managed Agents Overview -> Scaling Managed Agents: Decoupling The Brain From The Hands -> Agent Harness Design
• Codex path: Codex Operating Practices -> GitHub OpenAI Codex Repository -> Codex Best Practices
• Codex bridge path: Codex Plugin For Claude Code -> Codex Subagents -> Codex Operating Practices
• Codex skills path: Codex Agent Skills -> Codex LLM Wiki Skill -> Knowledge Operations
• OpenClaw operations path: OpenClaw Field Notes -> OpenClaw Field Note Capture Loop -> OpenClaw Field Note Template
• Claude Code path: Affaan Mustafa Claude Code Shorthand Guide -> Affaan Mustafa Claude Code Longform Guide -> Everything Claude Code GitHub Repository
• Operator-playbook path: Claude Code Best Practice Repository -> Claude Code Best Practice Tips Compendium -> Agent Harness Design
• Prompt-surface path: Claude Code System Prompts Repository -> How Claude Code Builds a System Prompt -> Codex Operating Practices
• Tooling-bundle path: Compound Engineering Plugin -> gstack -> Agent-First Repositories
• Managed-agent platform path: Multica -> Claude Managed Agents Overview -> Agent-First Repositories
• Ralph path: Ralph Wiggum Loop Technique -> Ralph Step-By-Step Guide -> Ralph GitHub Repository
• Design-systems path: Agent-Readable Design Systems -> VoltAgent Awesome DESIGN.md -> Codex Operating Practices
• Diagramming path: Architecture Diagram Generator -> Agent-Readable Design Systems -> Codex Operating Practices

Workflow

• Add immutable source files to raw/
• Integrate them into wiki/sources/ and the relevant long-lived pages
• Keep index.md, index.zh.md, and log.md current
• Revisit the corpus with lint passes so structure and synthesis improve over time