#codex

Reinstalling is easy. What's hard to get back on a new computer is your MCP servers, Skills, rules files, env vars and project dependencies. Ask your AI agent to inventory them, set up a private backup repo, list your keys, and write a restore script. Copy-paste prompt included.

An open-source Agent Skill that lets AI agents like Claude Code and Codex write instrumental music note by note, render it with real instrument samples, and export MP3, MIDI, stems and an editable project. Includes seven example tracks, setup requirements, commercial limits and a ready-to-use prompt.

A rumor says GPT-6.2 Astra launches Thursday, but the graphic itself is labeled as an unconfirmed concept. We checked OpenAI's news page and Codex's open-source code: there is no GPT-6.2, and Daybreak is a cybersecurity access program, not a new model. Don't change your setup over a rumor.

Beginner tutorial: Use the open-source Hypit to give Claude Code or Codex a video-making skill. Hand it a reference video, have it break down the opening hook, pacing, and subtitle structure first, then swap in your product, presenter, and language to regenerate. Get a price quote before spending money, and reuse assets to batch out several versions. Includes 4 copy-ready prompts.

Beginner tutorial: Use the free, open-source givemotion project to have Claude Code or Codex check your computer, set up the tools, and turn one paragraph of description into a promo video, animated data chart, or vertical short. Review an overview image before rendering to MP4, then change only the parts you name in plain language. One video can also be batch-exported into several versions. Includes 4 copy-ready prompts.

Beginner tutorial: turn one full-body image of your character into transparent animations with sprite-gen, an open-source Skill for Codex or Claude Code. Choose between GPT frame-by-frame or Grok video, write the actions, loops and export format, remove bad frames in the curator, and export animations for videos, animated stickers or websites. Includes 3 copy-paste prompts.

Use when the user wants to make 3D animation, 3D characters, cutaway demos, science maps and timelines, or to edit short videos or add sound effects. 12 Skills plus two commands, /3dg and /2dg, work with free tools such as Blender, Three.js, Remotion, and FFmpeg. Clarify needs first, make a preview before the final version, and get consent before installing any software. Use case: Let an agent make 3D animation, characters, science maps and short videos for you

Use when a user wants promotional material for a product, such as a homepage hero image, feature cards, a cursor-driven demo animation, or a Bento feature grid, and does not want to screenshot the real product. Start from a bundled example, build a faux UI from React components, record the full loop and run the automatic check before delivery. The output is an illustration and does not operate the real app. Video export needs a separate recording. Use case: Have an agent use one faux UI to make a demo animation, homepage hero image and Bento feature cards for your product

One faux UI produces a demo video, a homepage hero image and feature cards. The three assets share one style, and changing scenes does not require a new design.

Use when the user wants DOCX, PDF, TXT, XLSX, or business data turned into a dense, consulting-style PPT (industry research, executive briefings, board materials, client proposals). Based on the open-source CyberPPT (MIT): it first builds an evidence table and an SCR storyline, then picks one of 8 built-in styles and makes a page-by-page blueprint, and finally rebuilds each page as an editable PPTX with PptxGenJS and renders it for comparison. The user confirms all three stages. Use case: Have Codex turn your reports and data into a dense, consulting-style presentation that you can edit directly in PowerPoint.

Intermediate tutorial: For people who write analysis and client decks. Use Codex and the CyberPPT skill to build an editable, consulting-style PPT in three phases. Includes 3 copy-ready prompts.

A beginner tutorial: have your agent install Vercel's open-source agent-desktop, grant the accessibility permission yourself, then add the official Skill. Start with a read-only trial, have it describe its plan and wait for your go-ahead, and finally hand it the work of clicking back and forth between several apps, checking the interface when it's done. Includes 3 copy-ready prompts.

Use when the user doesn't know Git and you need to create, edit or write files in a folder or code, or when the user says save a snapshot, show what changed, revert, upload to GitHub, or merge. Before acting, save a snapshot first; start new features on a branch; after changes, report in plain language with how to undo; always ask before uploading, publishing, merging, or any action that could discard changes; secrets never go into the repository. Use case: Let the Agent manage every change for you when you don't know Git: snapshot before changes, explain after, roll back if something breaks, ask before uploading.

Use when a user wants the K-line structure, support and resistance, risk control plan, future volatility range, or backtest of an A-share, Hong Kong, or US stock, ETF, or index. Based on the open-source stock-skill (Apache-2.0): real daily market data, deterministic calculation, look-ahead-free backtesting, output as JSON, PNG, offline dashboard and Chinese report; only relays script numbers, outputs no probabilities, and stops if market data is unavailable. No fundamental analysis or investment advice. Use case: Have the Agent use real market data to compute a stock's K-line structure, support and resistance, and risk control levels, generate charts and reports, without fabricating numbers or stating probabilities.

Use when the user wants to turn a GitHub repo, technical article, PDF, local project or notes into an architecture or flow diagram, export an existing diagram as an image, video, GIF or Mermaid, or switch it to blueprint, hand-drawn, 3D and other styles. Based on the open-source arch-diagram-skill (MIT): reads the repo without running its code, draws inferred parts as dashed lines, and delivers only after self-check passes. Use case: Have the agent read an open-source project or an article, draw it as a moving architecture diagram, then switch styles for the occasion and export as an image or video.

A beginner's tutorial: get Node and a browser ready, install the arch-diagram Skill with one command, give a GitHub repo or article link to Claude Code or Codex to draw as a living architecture diagram, verify the inferred parts, switch between 8 styles with one sentence, and export as image, video, GIF, or Mermaid. Includes 2 copy-ready prompts.

Beginner tutorial: editonny is a free, open-source video editing skill. Transcription and rendering run on your own computer. Install it with one sentence, use plain language to trim pauses, remove slip-ups, add captions, switch to vertical format, and turn a passage into an animation. Includes how to read the settings card, default settings, and things to watch out for.

When the base-tier Claude quota runs out, it's usually because bulk output work was also handed to it. Keep judgment tasks for Claude, send bulk generation to Codex or a cheaper tool, and write the division of labor into CLAUDE.md.

Hand the tutorial to an Agent and it works through it to the end; you only confirm key steps such as logging in and paying, so learning and putting it into practice happen in one go.

Advanced tutorial: add claude_cli, codex_cli and fallback LLM providers to xiaozhi-esp32-server, calling the locally logged-in official CLIs, with all tokens handled by the CLI itself. Tool calls use a <tool_call> text protocol and are not spoken aloud. If Claude errors, it automatically switches to Codex, and the failing provider cools down for 300 seconds. Includes full code, configuration, testing methods, and measured latency.