Plans
-
plannotator.ai
Diagrams
- archify
- pretty cool
- JSON source of truth, which then generates the HTML
- playable (although that hasn't seem that useful)?
- lots of neat animations
TODO: diagram design skills?
Browser
Playwright
- screenshots and record videos
bunx playwright codegen https://example.com- manualy generate
TODO
- Vercel's agent browser?
- Ego browser? (lite.ego.app)
- some headless panda browser built in rust or something that's super small?
- wouldn't all these browsers have issues with rendering and loading sites properly if they're not Chromium?
GitHub
gh extension install drogers0/gh-image- Upload images and videos to PR comments
Problem: LLM creates PR comments that aren't associated to lines
- I installed Chrome Dev Tools MCP
claude plugin install chrome-devtools-mcpclaude mcp add chrome-devtools bunx chrome-devtools-mcp@latest
- TODO: need to fix login, right now, I'm logging in manually
- could set the manually
gh extension install drogers0/gh-image- to
Analyze the diff in the PR to identify all modified files containing UI components or CSS changes. For every visual code block altered:
1. Identify the local URL/route where that component renders on http://localhost/example/
2. Use Chrome DevTools MCP to navigate there and take a high-resolution element screenshot of that specific UI section.
3. Save each screenshot locally (e.g., component-name.png).
4. Post an individual inline review comment directly on the modified file and line(s) in the GitHub PR diff for each altered visual component. Each inline comment should contain:
- A short summary of the visual/UI change.
- The corresponding uploaded screenshot image.
Ignore purely logic-based files (API routes, utility functions, state management) that do not directly render HTML/CSS.