JC

Open source. 65 skills. Works across 10 AI coding tools. muse.mythslabs.ai

Agent governance is turning into its own category. Runtime policy engines. OWASP risk frameworks. Enterprise-grade interceptors. Big companies are shipping serious infra for this. I govern my agents with pure markdown files. The problem is the same though. Your agent forgets what it decided two sessions ago. Silently drops context. Fixes one thing, breaks two others. Other approach: intercept every action at runtime. Mine: make the agent write down what it did before it leaves. Pure markdown. No runtime, no database, zero dependencies. Over 9 versions the framework found 13 bugs in its own operating procedures. Diagnosed and patched each one by itself. One version the SOP hit 565 lines from all the patches. My agent started failing basic tasks because its own rulebook was too long. Had to cut 78% of it. Got more reliable immediately. Another version the agent wrote memory to the wrong directory for 4 days. No errors. Just silently saving to a path nobody reads. The part I really like is the framework keeps getting smarter about me. It tracks how I make decisions, not just what I decided. Different context depending on whether I'm doing strategy work or debugging code. Feels less like a tool and more like a second brain that actually remembers. More versions in. Still just markdown files.

Found a silent data loss bug in our own governance SOP today. Mixed-role conversations were only syncing one role, quietly dropping strategic decisions. Fixed it with a multi-role router + decision forward-write detector. Then ran Graphify on the whole codebase. 456 nodes, 43 communities. Now we can see the architecture before touching anything. The irony of a governance framework losing its own governance data is not lost on me.

Issue: github.com/myths-labs/mus… Release: github.com/myths-labs/mus…

Someone opened an issue asking for OpenCode support. Looked at it. OpenCode uses .agents/ (plural) instead of .agent/ (singular). That's it. Half a day later: install.sh, convert-skills.sh, docs, QA all done. v3.0.3 tagged. This is the part people don't talk about with "pure markdown, zero dependencies" It's not just about being lightweight. It's that adding a new tool literally means changing a directory path. No API adapters. No SDK updates. No CI pipeline changes. 7 tools supported now. The 8th won't be any different.

github.com/myths-labs/muse 65 open-source skills. Zero dependencies. Pure markdown. Works with Claude Code, Cursor, Copilot, OpenClaw any AI tool.

I spent a day inside Claude Code's leaked 512K lines of source code. Not for the architecture. Not for the agents. For 7 plain-text rules buried inside the TypeScript. Rules that solve problems every AI power-user hits: 1. Your AI remembers stale info from 2 weeks ago 👉 Tag every memory with a date. Flag anything >7 days. 2. Your memory file bloats to 500 lines of garbage 👉 Never save what you can re-derive. If you can grep it, don't store it. 3. You delegate to sub-agents but they have zero context 👉 Iron law: understand it yourself first, then write a self-contained spec. Never forward raw chat history. 4. "Last Thursday", which one? 👉 Always write absolute dates. 2026-04-01, not "Thursday." 5. Your AI only remembers mistakes, never what worked 👉 Record both corrections AND confirmations. 6. All memories dumped in one giant file 👉 4 categories: user prefs / feedback / project / reference. 7. End-of-session summaries are useless walls of text → Structured format: Rule → Why → How to apply. 💫 These aren't code. They're just rules that you can just copy them into your CLAUDE.md/AGENT.md We did, it's called MUSE v3.0

I was being sarcastic when I said the @antigravity Ultra plan would be consumed within 20 minutes using Opus at this rate. I didn't expect it to actually become a fulfilled prophecy. What a shame. @GeminiApp

Shipped 3 things before lunch: 1. Published an MCP server to npm. Any AI agent can plug into our platform with one command now. 2. Built a dynamic routing layer that detects asset types and auto-switches backend providers. 3. Fixed a button that literally had no onClick handler. Stared at it for 2 days before realizing it was just dead markup. The MCP piece is the one I'm most excited about. Protocol-level distribution > marketing.

Testing a bidirectional radar for @Airachne today. Teaching my digital twin how to understand and organize my mentions in real-time. If you get a reply from this account slightly faster than my usual human speed, the test is probably working.

Spent hours debugging a background running issue today. Finally fixed it and the new architecture works great! Next step: building a Mission Control UI. It's time to test the water.

@lianyanshe 我自己每天在用

@SunshiningDay 太全面了

非常感谢大家的推荐 Claude code skill，我整理一下评论区推荐的，装上大部分了，这几天正好体验一下。 有想要安装同款 skill 可以找到 github 对应的仓库，把链接给 Claude code 即可，新手推荐先把CodePilot+Superpowers+Planning-with-files+Context7用起来就足够了 下面是大家精选的和我自己用起来觉得不错的 1. Superpowers (强烈推荐) GitHub：obra/superpowers 工程化能力很强，适合功能开发、系统化调试、头脑风暴、子代理协作、代码审查、git worktrees 工作流。是一整套让 Claude Code 按软件工程流程做事的方法论。 2. CodePilot (强烈推荐) GitHub：op7418/CodePilot 一个给 Claude Code 用的桌面 GUI 客户端，适合不想只用命令行、图形化管理 MCP 和 skills、想做项目级 workspace 和记忆的人。需要先安装Claude Code才能使用 3. PUA GitHub：tanweai/pua 真的不是在搞抽象，确实有用。解决 AI反复失败，甩锅、停止验证、只给建议不落地时使用。这个 skill 的核心就是把 AI 拉回到更强执行、更强排查、更少放弃的模式，还支持手动 /pua 触发。 4. Planning-with-files GitHub：OthmanAdi/planning-with-files 像 Manus 那样工作，非常适合多步任务、研究任务、大改造、长链路实现。它的核心思路是把计划、发现、进展写进 markdown 文件，让 Claude 不要只靠上下文硬记，从而降低目标漂移和重复犯错。 5. skill-creator GitHub：anthropics/skills/skills/skill-creator Anthropic 官方仓库里的样例和工作流模板，提供初始化、校验、打包整套能力。 6. memory-bank GitHub：SametEge/claude-memory-bank 适合你经常中断会话、隔天继续、项目信息容易散掉的场景。它会维护结构化记忆文件，让 Claude 记住架构决策、剩余任务、上下文历史。 但对于多数人来说把项目初始化 /init 和 CLAUDE.md 用好，就已经够用了。只有你明确需要更结构化的项目记忆，再上 memory-bank。 7. Context7 GitHub：upstash/context7 给 Claude Code 提供最新代码文档和 API 资料，减少过期知识、错 API、旧版本写法这类问题。会走 OAuth、生成 API key、安装对应 skill。 8. oh-my-claudecode GitHub：Yeachan-Heo/oh-my-claudecode Claude Code 的多代理编排增强包，主打零学习成本、装完就能用。它把 Team、多代理流水线、autopilot、deep interview、skill 管理这些能力整合到一起，适合想让 Claude Code 更自动、更会拆任务的人。 还有GitHub、git、notebookLM、find等 skill，就不一一介绍介绍了，各位大神还有好的也欢迎在评论区推荐，非常感谢

推荐一个：myths-labs/muse 跟大部分 Skill 定位不同，它不是解决某一件事的 Skill，是一整套 AI 编程工作操作系统，把记忆、技能、分工串到一起。 几个单独装 Skill 做不到的事： - 纯 Markdown 零 npm 依赖，30 秒寄生到任何项目 - /resume 5 秒恢复跨对话上下文（不用每次开新对话重贴 200 行背景） - /distill 自动蒸馏长期教训（犯过的错永远不会再犯） - /bye 一键存档，下次自动接上 - 记忆自动过期 + 新旧自动关联，不会被过时信息污染 - 角色隔离：开发/测试/增长/战略各管各的，AI 不会在 debug 时建议你做营销 - 58 个内置 Skill + MCP Server + VS Code Dashboard - 兼容 OpenClaw / Claude Code / Cursor / Windsurf / Antigravity / Codex 等 6 种工具 简单说就是：一般 Skill 是"给 AI 加一个技能"，MUSE 是"给 AI 一套有记忆有纪律的工作操作系统"。 GitHub: github.com/myths-labs/muse

如果让你推荐 3 个 Claude code 必装插件或者 skill 你会选哪三个，小弟真心求教！

@antigravity How about pay some attention to what really is going on huh? discuss.ai.google.dev/t/ultra-dramat…

Antigravity is more than just a tool. It’s a platform you use to navigate the logic and the execution in one place. By leveraging Gemini’s multimodal power, we’re enabling everyone to think through complex problems and build directly upon those insights, creating things they never dreamed possible.

Google is silently cutting Ultra plan quotas by half every single day? I'm paying $250/month and using @antigravity as my daily coding pair. Few weeks ago: 5-6 parallel agents, 5+ hrs, no issues A week ago: same setup → quota hit at 4 hrs 3 days ago → 2 hrs Yesterday → 2 hrs with just 1-2 conversations Today → ~45 mins 👎👎👎 At this rate I'll probably get less than 20 minutes tomorrow running a single conversation. $250/month, for ~45 mins/5 hrs. No announcement. No email. No changelog. Nothing. @Google's only response in their @googledevs forum? A marketing paragraph about "seamless scaling". Totally unacceptable. @GeminiApp discuss.ai.google.dev/t/ultra-dramat…