$ loading_
35948 Skills & MCP tools — copy a prompt, zero pre-install, into Claude Code / Codex / Cursor.
🎓 Skill = teach the AI how to do a task, copy & use; 🔌 MCP = connect the AI to an external tool/data source, usually needs a key.
Demonstrate MCP tools, resources, and prompts for comprehensive protocol capability testing.
Helps AI break down complex problems step by step and reflect on its reasoning.
Store and recall cross-conversation details with persistent knowledge-graph-based memory.
Fetch web pages and convert them to Markdown for AI reading and analysis.
Write, understand, and debug code in the Lobster programming language.
Chat directly with the Discord-backed OpenClaw agent in real time.
Summarize CodexBar local cost logs by model for Codex or Claude.
Diagnose OpenClaw issues by choosing logs, probes, and proof paths first.
Transcribe audio into text or speaker-separated transcripts with OpenAI APIs.
Delegate coding tasks to background AI coding agents for implementation and changes.
An AI coding assistant for review, debugging, refactoring, and design checks.
Convert text to speech locally and offline with sherpa-onnx, no cloud needed.
Route coding-agent requests into the right ACP or acpx session flow.
Create, refine, validate, and restructure AgentSkills and SKILL.md files.
Transcribe audio to text locally with Whisper CLI, no API key required.
Convert text into high-quality speech with a macOS say-like workflow.
Manage, authenticate, call, and inspect MCP servers and tools over HTTP or stdio.
Maintain an OpenClaw memory wiki with deterministic, source-backed page updates.
Use Gemini for one-shot prompting, summaries, generation, coding, and tool routing.
Create and review technical docs and agent instruction files in repositories.
Search and analyze past session logs with jq for context and insights.
Set conversation rules to discover and invoke skills before replying.
Create, revise, and validate AI skills before deployment.
Force factual investigation before edits or commands to improve output quality.
从访谈内容留存、转写回看，到主题归纳与产品建议，一套串起用户研究整理流程。
从流量采集、营销归因到数仓查询与看板输出，一套串起自动取数和分析闭环。
从读取设计稿、提取上下文到映射组件库与同步评审，这套组合很适合设计到前端落地的协作链路。
In-depth guides on Skill / MCP setup and security — beyond the directory listings
普通聊天 AI 很难真正读懂大型仓库，因为它缺少结构、上下文和可验证的检索路径。本文从 0 到 1 搭一个只面向本地代码库的 Cursor 助手：能查结构、追调用链、看 diff、给重构建议，还能尽量压住幻觉。
Agent-hop 是一个基于 Rust 的 TUI，支持在进行中的 Claude Code 聊天里切换到其他编程代理，尽量不丢失上下文。项目作者提到，它最早源自一个简历与搜索工具，后来因为使用过程中的阻碍，单独做成了现在的界面。
Harden.run 表示，针对能够生成任意代码的 Coding Agent，单靠传统红队测试并不够。他们通过后训练网络安全小模型、调整推理方式，并结合 inline reference monitoring 等程序分析手段，在 LinuxArena 和 SleightBench 上取得了超过 GPT5.5-xhigh 的成绩。