Agent Skills

browser-video-recording

Create polished 60 fps 4:3 4K browser screen-recording style videos from Codex in-app browser captures, with browser-only crop, natural macOS cursor styling, deliberate click choreography, zoom-follow framing, ffprobe/thumbnail verification, and optional native recording compatibility checks. Use when the user asks to record or re-record browser actions, show cursor clicks and zooms, make Dribbble/UI inspiration or product demo recordings, or asks whether Codex, Playwright, or an MCP can produce

Install

npx skills add https://github.com/mengto/skills --skill browser-video-recording
SKILL.md

Browser Video Recording

Dependency Model

Separate the workflow into three layers:

  1. Browser control and source frames: Use the Codex in-app browser through the Browser skill/MCP. Do not use Chrome when the user asks for Codex browser or the project says not to. The browser layer supplies real screenshots and real click/scroll/navigation states.
  2. Local video renderer: Use the bundled Python script to render screenshots into an MP4 with a natural macOS cursor, subtle click scale, and calm zoom-follow framing. This does not require a screen-recording MCP.
  3. Optional native recorder integration: Use a native recording MCP only when callable and explicitly useful for starting/stopping an actual app recording or opening a generated project. If unavailable, blocked, or unable to capture the Codex browser surface, fall back to the screenshot-to-video renderer and explain the fallback.

Self-Contained Execution

This skill is enough for another Codex instance to perform the Python rendering as long as the full skill folder is installed, including scripts/render_browser_demo.py. Codex should run the bundled script instead of recreating the renderer.

Required local tools:

  • python3
  • Python package Pillow
  • ffmpeg and ffprobe

Optional macOS cursor extraction uses swift and /usr/sbin/screencapture. If those are unavailable, provide a transparent PNG through cursor_asset and set cursor_hotspot in the config.

The Browser skill/MCP is still needed for fresh Codex in-app browser screenshots. If it is unavailable, use the best available browser automation screenshot source and state the fallback.

Optional Native Recorder Support

For native app recording, verify the MCP supports:

  • Use frameRate: 60.
  • Use resolutionRawValue: "4K".
  • Use ratioPresetRawValue: "4:3" for region recordings.
  • Use an explicit selectedRegionOnScreen when cropping to an embedded browser viewport.

For the local renderer, 4K 4:3 means 3840x2880. If a native recorder exposes a 4K enum, treat it as an app-side resolution preset and pair it with 4:3 region selection when recording through that MCP.

Capture Workflow

Use browser:control-in-app-browser first. After loading its docs, drive the in-app browser with the Node/browser API:

  • Set viewport to 1920x1440 for 4:3 4K output, then reset it before finishing.
  • Navigate to the requested page and perform actual clicks, backs, searches, and scrolls.
  • Save browser-only screenshots after each important state change. Use names like 01-results-top.png, 02-after-card-click.png, 03-back-results.png.
  • Record the viewport coordinates of each click/scroll target. The final cursor path should land on those points, pause, click, then move to the next meaningful point.

Avoid desktop coordinate automation unless the user explicitly needs system UI. Codex browser visibility may not expose a capturable macOS surface; screenshots from the browser API are the reliable source.

Motion Rules

Make the cursor feel purposeful:

  • Use sparse, mostly straight cursor paths.
  • Pause briefly on a target before clicking.
  • Do not wander over unrelated UI.
  • Let the page transition happen after the click, not before.
  • Use only subtle click feedback: scale dip/rebound, not decorative rings unless requested.
  • Keep zoom-follow calmer than the cursor. The camera should frame the clicked content, not chase every small mouse move.

Default visual target:

  • Real macOS pointer extracted from the compositor, or an existing transparent cursor asset.
  • Cursor scale 3.0, matching the large cursor style commonly used in edited demo recordings.
  • Click animation strength 15 for a subtle click scale dip.
  • Rotation strength up to 8deg, only from horizontal velocity.

Render Script

Use scripts/render_browser_demo.py for deterministic 60 fps 4:3 4K output:

python3 /path/to/browser-video-recording/scripts/render_browser_demo.py \
  --config /tmp/browser-demo-config.json \
  --output /tmp/browser-demo-4x3-4k-60fps.mp4

Generate a starter config:

python3 /path/to/browser-video-recording/scripts/render_browser_demo.py --write-template

The config supplies:

  • shots: map of state names to screenshot PNG paths.
  • scene_starts: [time, shotName, transition] entries.
  • cursor_keys: [time, x, y] viewport-coordinate entries.
  • camera_keys: [time, x, y, zoom] entries.
  • click_times: click timestamps.
  • duration, fps, output_size, source_size, cursor_scale.

Preferred defaults:

{
  "fps": 60,
  "output_size": [3840, 2880],
  "source_size": [1920, 1440]
}

If cursor_asset is omitted, the script briefly shows a solid-color overlay and uses screencapture to extract the real current macOS cursor into a transparent PNG. This is local rendering support, not an MCP dependency.

Verification

Always verify the delivered video:

ffprobe -v error \
  -show_entries stream=codec_name,width,height,r_frame_rate \
  -show_entries format=duration,size \
  -of json /path/to/output.mp4

Extract at least one mid-video thumbnail and inspect it:

ffmpeg -y -ss 8 -i /path/to/output.mp4 -frames:v 1 -update 1 /tmp/browser-demo-thumb.png

Confirm:

  • Resolution is the requested output, usually 3840x2880 for 4:3 4K.
  • Frame rate is 60/1 or equivalent.
  • Duration matches the request.
  • Cursor is visible, natural, and on-target.
  • The video is cropped to browser content, not the desktop.
  • The browser viewport override has been reset.

Related skills

video-editgenmedia-labs715KEdit existing video on RunComfy — this skill is a smart router that matches the user's intent to the right edit model in the RunComfy catalog. Picks Wan 2.7 Edit-Video (general restyle / background swap / packaging swap, identity + motion preservation), Kling 2.6 Pro Motion Control (transfer precise motion from a reference video to a target character), or Lucy Edit Restyle (lightweight identity-stable restyle / outfit swap). Bundles each model's documented prompting patterns so the skill gets shai-video-generationgenmedia-labs714KGenerate AI videos on RunComfy via the `runcomfy` CLI — a smart router across the full video-model catalog: HappyHorse 1.0 (Arena #1, native in-pass audio), Wan-AI Wan 2-7 (open weights, audio-driven lip-sync), ByteDance Seedance v2 / 1-5 / 1-0 (multi-modal cinematic), Kling 3.0 / 2-6, Google Veo 3-1, MiniMax Hailuo 2-3, ByteDance Dreamina 3-0. Covers text-to-video (t2v), image-to-video (i2v), and Veo's video-extend endpoint. The skill picks the right model for the user's intent (Arena-#1 qualitai-musicgenmedia-labs714KGenerate AI music on RunComfy via the `runcomfy` CLI — a smart router across the music-model catalog. Routes to ElevenLabs AI Music Generation (premium 44.1 kHz stereo vocal tracks, 5 s–5 min, $0.0083/s) and ACE Step / ACE Step 1.5 (StepFun-AI open-weights, tag-driven composition, multilingual lyrics, $0.0002–0.0003/s, ~27× cheaper), plus ACE Step audio-inpaint (regenerate a time range inside an existing track) and ACE Step audio-outpaint (extend a track before or after). Picks the right model fimage-to-videogenmedia-labs713KAnimate any still image on RunComfy — this skill is a smart router that matches the user's intent to the right i2v model in the RunComfy catalog. Picks HappyHorse 1.0 I2V (Arena #1, native audio, identity preservation) for general animations, Wan 2.7 with `audio_url` for custom-voiceover lip-sync, or Seedance 2.0 Pro for multi-modal animation from image + reference video + reference audio. Bundles each model's documented prompting patterns so the caller gets sharper output without burning iterat

Search skills and MCP servers

Fuzzy search across 23,137 skills and servers