claude-ui-feedback

A Claude Code skill that turns a screen recording + feedback into an annotated HTML report — with red-boxed screenshots highlighting exactly what was discussed.

The HTML output is self-contained (images embedded) and imports directly into Notion. A plain-text summary is also copied to your clipboard for easy pasting into tickets or PRs.

What it does

1. Extracts frames from the recording
2. Transcribes audio feedback via Whisper, or uses written feedback you provide
3. Matches feedback to the relevant frames and draws red annotation boxes
4. Smartly chooses between stacked (multiple boxes on one frame) or individual annotated images
5. Assembles everything into a single .html file in your current directory
6. Copies a feedback summary to your clipboard

Install

One-liner:

curl -fsSL https://raw.githubusercontent.com/nomativ/claude-ui-feedback/main/install.sh | bash

Or clone first:

git clone https://github.com/nomativ/claude-ui-feedback.git && bash claude-ui-feedback/install.sh

The installer will:

1. Copy the skill and helper script to ~/.claude/
2. Check for required dependencies (ffmpeg, Pillow, openai-whisper) and offer to install them
3. Remind you to restart Claude Code

After restarting, /ui-feedback will appear automatically.

Usage

/ui-feedback path/to/recording.mov

Or with written feedback instead of audio:

/ui-feedback path/to/recording.mov "the save button is hard to find and the modal closes too fast"

Output lands in your current working directory as <video-name>-feedback.html.

Annotation modes

The skill automatically picks the best annotation style:

• Stacked — multiple boxes layered on a single frame. Used when all feedback targets the same screen and there are 3+ annotations.
• Individual — one annotated image per feedback point. Used when feedback spans different screens or there are fewer annotations.

Import into Notion

After running the skill, open Notion → Settings → Import → HTML and select the generated .html file.

Uninstall

rm -rf ~/.claude/skills/ui-feedback ~/.claude/scripts/ui_feedback.py