Repository navigation
Browser automation integration for post-implementation testing #15
Description
Summary
sfk run --browser
sfk run "test the login flow" --browserGive the AI access to browser automation for testing web UIs after implementation.
Why
AI can write code but cant verify it visually. Browser automation closes the loop — implement a feature, then actually test it in a browser.
Implementation
Inject browser instructions into the prompt when --browser flag is set. The AI uses the browser tool via shell commands. No sfk code changes needed beyond prompt injection.
Supported tools (detect whichever is installed):
agent-browser— snapshot-based, element refsplaywright— script-based- Or just
curl/web_fetchfor API testing
## Browser Testing (enabled)
After implementing, verify your changes:
- agent-browser open http://localhost:3000
- agent-browser snapshot
- agent-browser click @e1
- agent-browser screenshot verify.png
This is literally just appending text to the prompt. The AI and Claude Code handle the rest.
Activity