Screen Recording — Design skill for Claude Code

screen-recording

**An agent skill for script-driven product capture.** Works with Claude Code, Codex and Cursor — a plain `SKILL.md` skill plus one Python script, nothing agent-specific inside.

Give your coding agent a shot list and it records clean product footage — a native app in the iOS Simulator, a website in a driven browser, or a Figma prototype — using a staging checklist that burns one-time dialogs off-camera, a per-take protocol that verifies every end frame, and an auto-zoom derivation step that turns *measured* interaction times into a camera plan and cursor cues. The one rule everything hangs on: **trim from measured times, never from planned ones** — tap-injection latency makes a 15-second plan into a 60-second take.

What's inside

Path	What
SKILL.md	Triggers, the capture loop, non-negotiables
references/capture-protocol.md	Surface selection, staging checklists (sim / web / Figma proto), take protocol, HiDPI resolution maths, failure modes
scripts/autoplan.py	Measured tap times → auto-zoom camera plan + cursor/ripple arguments

Requirements

• macOS. iOS Simulator capture needs Xcode (DEVELOPER_DIR=/Applications/Xcode.app/Contents/Developer); website/Figma capture needs screencapture (built in) and one-time Screen Recording permission for your terminal.
• python3 for autoplan.
• An agent that reads SKILL.md skills — Claude Code, Codex, or Cursor.

Install

**curl** — installs for every supported agent (one real copy, the rest symlinked):

curl -fsSL https://raw.githubusercontent.com/Malik1942/screen-recording/main/install.sh | bash

Narrow to specific agents or scope to one repo:

./install.sh --claude --cursor

./install.sh --project

**git:**

git clone https://github.com/Malik1942/screen-recording.git ~/.claude/skills/screen-recording

| Agent | Pe