brag

brag is an Agent Skill open-sourced by Shunit Haviv Hakimi (the repo sits under the GitHub account latent-spaces) that makes a launch video for a website or app you've built. Say /brag to your AI coding agent inside the project and it reads the code first: what the product does, the one line worth bragging about, the most camera-ready part of the interface, and the two or three steps a user takes. It writes a plan and a storyboard, then hands a brief to the video composition tool Hyperframes to lay out, time and render. The video runs 15 to 25 seconds by default, in landscape, vertical or square, with seven tone presets such as polished, a fake startup launch and deadpan, or a tone you describe in your own words; music and sound effects are on by default, and narration only turns on with --voice, using Kokoro. You end up with a brag-output folder holding brag.mp4, a chosen poster frame, the plan and share copy ready to post. The repo has a second Skill, brag-slim, which skips Hyperframes and the bundled assets and has the model build the visuals, music and sound effects with whatever tools are on the machine; besides a local project, you can give it a website URL. The README says brag-slim was rebuilt lean for Claude Opus 5.5, and on Opus 5.5 /brag switches to it automatically unless you add --full. What sets it apart is that it takes its material straight from your source code, with no live URL or screenshots needed, and its rules require at least one shot of the product's real interface or copy.

Who it's for:developers and indie makers who've built a site, tool or app with AI and want something to post so people see it. You'll need to be comfortable with an AI coding tool that runs commands, such as Claude Code or Codex. For readers who don't code and only want the video, the README points to letsbrag.app, which makes one from a pasted link; that is a separate online service AI Inspiration World hasn't checked.

Before you start: Claude Code and Codex can install it from the plugin marketplace, and the plugin includes both Skills; other tools can install either one with npx skills add plus --skill brag or --skill brag-slim. The full version needs Node.js 22 or later, FFmpeg and the Hyperframes CLI, and the beat-analysis script uses Python and uv. It reads your project's code, README and images; given a URL, brag-slim goes online to open the site, using a headless browser when needed, and downloads the images and videos it will use into the output folder. SKILL.md tells it not to read .env, key or credential files, and not to carry secrets, internal URLs, real customer names or email addresses into the video, swapping in fictional stand-ins where the interface shows them; if your project holds client data or unreleased work, check it yourself before running. Share copy is written to a file and nothing is posted for you. The bundled sound effects are CC0 from Kenney and the music comes from ende.app; the repo's own note says the exact music license terms still need to be verified and documented, so confirm them before publishing a video with that music. The code is MIT licensed; the AI model and dependencies such as Hyperframes come with their own costs and terms.

Based on the project's public documentation. Not tested by AI Inspiration World.