发现优质的 AI Agent 技能

You are an expert accessibility auditor specializing in WCAG 2.1 Level AA compliance.

"Persistent, searchable, context-aware memory for AI agents. Store facts, preferences, corrections, and procedures across sessions and projects. Inject memory context directly into prompts. Zero dependencies, local-only, MIT licensed."

**Persistent, searchable, context-aware memory for AI agents. Store what matters. Never lose context again.**

"Learn from errors and corrections in real-time. Capture failures, user feedback, and successful patterns to continuously improve performance. Free and open-source."

**Learn from errors and corrections in real-time. Continuously improve by capturing failures, user feedback, and successful patterns.**

> **Purpose**: Record core info that rarely changes. Agents read on-demand, not every time.

来自 openclaw/skills 技能

> **For Claude Code** | 适用于 Claude Code > > **Note**: This methodology works with any multi-agent system (Cursor, LangChain, OpenAI Assistants, OpenClaw, etc.). This file is the Claude Code skill wrapper. See [README.md](README.md) for platform-agnostic usage. > > **说明**：本方法论适用于任何多智能体系统（Cursor、LangChain、OpenAI Assistants、OpenClaw 等）。本文件是 Claude Code 技能封装。通用用法见 [README.md](README.md)。

Add a trust gate in front of outgoing agent payments.

Browser control for AI agents via HTTP API. Workflow: navigate, snapshot, act.

Give your OpenClaw agents persistent, structured memory across sessions.

What is being created in this sprint must be a thoroughly detailed instruction set that can be followed with **zero context** every time. No step may rely on prior chat memory, implied intent, or ambiguous wording. Every directive must be executable as written by a cold-start agent.