| name | qa:tools |
| user-invocable | true |
| description | QA inspection tools for visual testing -- video frame extraction, screenshot analysis |
QA Tools
Tools for inspecting visual artifacts during QA workflows. Use this skill when you need to analyze screen recordings, identify UI regressions, or document visual behavior from test runs.
When to Use
- Reviewing screen recordings captured during QA or persona testing
- Investigating visual regressions flagged in bug reports
- Extracting key moments from UI walkthrough recordings
- Building visual evidence for bug reports or PR reviews
Video Frame Extraction
The video_frames.py tool extracts frames from a video file at regular intervals, saving them as images that can be inspected with the Read tool.
Invocation
Run from this skill's own directory (shown as "Base directory for this skill" when the skill loads):
cd <base directory of this skill> && uv sync && uv run video_frames.py <video_path>
Options
| Flag | Default | Description |
|---|
--output-dir | /tmp/qa-frames/ | Directory for extracted frames |
--interval | 2 | Seconds between frames |
--format | png | Output format: png, jpg, or webp |
Examples
Extract frames every 2 seconds (default):
cd <base directory of this skill> && uv sync && uv run video_frames.py /tmp/recording.webm
Extract frames every half-second for detailed inspection:
cd <base directory of this skill> && uv sync && uv run video_frames.py /tmp/recording.webm --interval 0.5
Save as JPEG to a specific directory:
cd <base directory of this skill> && uv sync && uv run video_frames.py /tmp/recording.webm --output-dir /tmp/bug-123/ --format jpg
Viewing Extracted Frames
After extraction, use the Read tool to view individual frames. The Read tool supports image files natively:
Read /tmp/qa-frames/frame_0000s.png
Read /tmp/qa-frames/frame_0004s.png
Workflow: QA Bug Report
- Extract frames from the screen recording attached to the bug report
- Scan the frames at a coarse interval (every 2-5s) to identify where the issue appears
- Re-extract at a finer interval around the problem timestamp if needed
- Read specific frames to examine the visual state in detail
- Document findings with frame references and timestamps
Tips
- Start with a coarse interval (2-5s) to get an overview, then narrow down
- Frame filenames include the timestamp (
frame_0002s.png = 2 seconds in), making it easy to correlate with video playback
- For long recordings, use
--output-dir to keep frames organized per investigation
- The tool prints each saved path to stdout, so you can pipe or parse the output
- Use
png for lossless quality when inspecting UI details; use jpg for smaller files when reviewing many frames
