AI Repo: calesthio/OpenMontage

Skyler: Hey everyone, welcome back to Git News A I, the show where we dig through the hottest trending repositories on GitHub so you don't have to. Before we jump in, huge shoutout to our sponsor LimitLess A I — they're doing some seriously cool stuff with personalized A I and we'll talk more about them at the end. Today we've got a wild lineup — leaked system prompts from basically every major A I company, a video production system run by agents, Chrome DevTools for your A I coding assistants, and more. Let's get into it. Skyler: Alright, kicking things off with something that genuinely made me do a double take — it's called OpenMontage, from the repo calesthio/OpenMontage, and it's sitting at over 33,000 stars already. The description calls it the world's first open-source agentic video production system, and honestly, that's not just marketing fluff. This thing is massive. Skyler: So what exactly is it? Think of it as turning your A I coding assistant — or really any agent-capable tool — into a full-blown video production studio. We're talking 12 distinct pipelines, 52 tools, and over 500 agent skills baked in. That is an absurd amount of capability packed into one system. Skyler: The idea is that instead of you manually stitching together video workflows, the agents handle the orchestration. Need a voiceover? There are T T S providers built in. Need to sequence clips, add effects, manage assets? The pipelines handle that. It's designed so that the agent knows what tool to reach for at each stage of production. Skyler: And speaking of T T S, the most recent commit on this repo is a DashScope integration for the T T S provider — DashScope being Alibaba Cloud's A I service. So the team is actively expanding the number of voice and speech options you can plug in. That tells you this project is moving fast and the ecosystem around it is growing. Skyler: What I find really compelling here is the democratization angle. Professional video production has always required either expensive software, expensive people, or both. An open-source agentic system that can handle the heavy lifting? That opens the door for indie developers, content creators, small studios — anyone who wants to produce high-quality video without a massive budget. Skyler: If you're building on top of agent frameworks, experimenting with multi-step pipelines, or just curious what it looks like when you give an A I system 500-plus skills to work with, this is absolutely worth exploring. It's written in Python, so the barrier to entry is low if you're already in the M L or A I space. Check the show notes for the link. Skyler: Next up, and this one is moving at a breakneck pace — hesreallyhim/awesome-claude-code, sitting at nearly 48,000 stars. Now, the name says it all — this is a curated collection of resources, tools, plugins, and skills specifically for Claude Code, Anthropic's coding agent. Skyler: If you haven't used Claude Code yet, it's Anthropic's A I-powered coding companion that lives in your terminal and can read your codebase, write code, run commands, and generally act as a seriously capable engineering collaborator. It's gotten a lot of traction in the developer community, and this repo is basically the community's knowledge base for getting the most out of it. Skyler: What's in here? We're talking hand-picked skills, status line configurations, agent setups, developer tooling integrations, and plugins. The repo organizer clearly has strong opinions about quality — the description is wonderfully over-the-top, calling it a showcase of quote top tier skills and scintillating status lines, which honestly made me smile. Skyler: The recent commits are automated ticker data updates, which tells you this repo is actively maintained with some automation around keeping everything fresh and up to date. That's a good sign for longevity. Skyler: Who should be bookmarking this? If you're using Claude Code daily, or even if you just want to get started and want to hit the ground running with battle-tested configurations, this is your go-to resource. It's also a great study in how the community builds around A I coding tools — the ecosystem that grows up around these agents is becoming as important as the agents themselves. Skyler: There's also something interesting here from a meta perspective — as A I coding assistants get more capable, the value of curated community knowledge about how to use them well goes way up. This repo is basically that knowledge crystallized. Definitely check the show notes for the link. Skyler: Okay, third repo and this one is super practical for anyone building or using A I coding agents — ChromeDevTools/chrome-devtools-mcp, clocking in at over 45,000 stars. The description is refreshingly simple: Chrome DevTools for coding agents. But there's a lot packed into that sentence. Skyler: M C P stands for Model Context Protocol, which is an open standard that lets A I models connect to external tools and data sources in a structured way. So what this project does is expose Chrome DevTools capabilities as M C P tools that your A I coding agent can actually use. Skyler: Think about what that means in practice. Your agent can now inspect the D O M, check console errors, analyze network requests, profile performance — all the things you'd normally do yourself in Chrome DevTools, but now your A I assistant can do them autonomously while it's working on your frontend code. Skyler: The most recent release is version 1.5.0, and there's a really interesting new tool that just landed — get_heapsnapshot_duplicate_strings. That is a very specific, very useful capability. Heap snapshot analysis is something developers use to find memory leaks and inefficiencies, and having an agent that can automatically flag duplicate strings in memory is the kind of thing that saves you hours of debugging. Skyler: This is coming from the actual Chrome DevTools team, which gives it serious credibility. This isn't a third-party hack — it's the official team thinking about how their tools should work in an agentic world. That's a really significant signal about where browser tooling is heading. Skyler: If you're building frontend applications and using any A I coding tools that support M C P — Cursor, Claude Code, Cline, take your pick — this is something you should be integrating. It dramatically expands what your agent can perceive and act on when working on web projects. Check the show notes for the link. Skyler: Now this next one — buckle up — because it's either the most fascinating or the most controversial repo depending on who you ask. asgeirtj/system_prompts_leaks has nearly 49,000 stars and it's exactly what it sounds like. It's a collection of extracted system prompts from basically every major A I company you can think of. Skyler: We're talking Claude models from Anthropic — Fable 5, Opus 4.8, Claude Code, Claude Design. OpenAI's ChatGPT 5.5 Thinking, G P T 5.5 Instant, Codex. Google's Gemini 3.5 Flash and 3.1 Pro. xAI's Grok. Cursor, Copilot, VS Code, Perplexity. This list goes on and on, and the repo description says it's updated regularly. Skyler: Now, why does this matter technically? System prompts are the instructions that shape how an A I model behaves — they define the persona, the constraints, the capabilities, the tone, and the decision-making logic of the product you're interacting with. These are usually closely guarded trade secrets, so having them out in the open is remarkable from a research and transparency standpoint. Skyler: For developers, this is genuinely educational material. You can study how Anthropic instructs Claude to handle edge cases, how OpenAI structures reasoning prompts, how these companies think about safety boundaries, capability constraints, and user experience at the prompt engineering level. There is so much you can learn here about professional-grade prompt design. Skyler: The recent commits are adding things like artifact design bundled skills and code review effort levels, which suggests this isn't just a static dump — someone is actively parsing and organizing these into usable components. The skills and dataviz additions are particularly interesting for anyone building their own agent systems. Skyler: There's obviously an ethical conversation to be had about how these were obtained and whether sharing them is appropriate — that's real and worth thinking about. But from a pure knowledge and transparency perspective, seeing what's actually under the hood of the products shaping how millions of people work with A I is genuinely valuable. Worth a look — check the show notes for the link. Skyler: And rounding out today's episode, we've got tirth8205/code-review-graph, sitting at about 19,000 stars, and this one is solving a really specific and really annoying problem that anyone working with A I coding tools on large repos has definitely run into. Skyler: The problem is context. When you ask an A I assistant to review code or help with a task in a large codebase, it often ends up reading way more files than it actually needs to. That burns through your context window, slows things down, and can actually make the quality of responses worse because you're flooding the model with noise. Skyler: Code Review Graph tackles this by building a persistent, local-first intelligence graph of your codebase. It maps out the relationships between files, functions, and dependencies, so when your A I tool needs to understand something, it can pull only the relevant parts of the graph instead of brute-forcing its way through your entire repo. Skyler: The key word here is persistent — this graph is built once and updated incrementally, so you're not rebuilding it from scratch every time. And local-first means your code stays on your machine, which is huge for anyone working on proprietary codebases who can't just ship everything to a cloud service. Skyler: It works with both M C P and C L I interfaces, so you can plug it into your existing A I coding tools or use it directly from the command line. And the team has benchmarked the context reductions, which I love — they're not just claiming it works, they're showing you the numbers on large-repo workflows. Skyler: There haven't been recent commits so it might be in a stable phase or taking a breather, but at 19,000 stars the community has clearly validated the concept. If you're doing serious development work on large codebases with A I tools, this is exactly the kind of infrastructure layer that makes everything else work better. Check the show notes for the link. Skyler: And that's a wrap on today's episode of Git News A I! Huge thank you to LimitLess A I for sponsoring the show — if you're building in the A I space, definitely check them out, link is in the show notes. If you found today's repos useful, please subscribe wherever you get your podcasts, leave a review if you're feeling generous, and as always, all the links to every repo we covered today are waiting for you in the show notes. We'll see you next time.