dogfood
Systematically explore and test a web application to find bugs, UX issues, and other problems. Use when asked to "dogfood", "QA", "exploratory test", "find issues", "bug hunt", "test this app/site/platform", or review the quality of a web application. Produces a structured report with full reproduction evidence -- step-by-step screenshots, repro videos, and detailed repro steps for every issue -- so findings can be handed directly to the responsible teams.
87
85%
Does it follow best practices?
Impact
87%
1.97xAverage score across 3 eval scenarios
Risky
Do not use without reviewing
Evaluation results
100%
34%QA Pass on a Demo Web Application
Session setup and command usage
Direct binary used
100%
100%
Screenshots subdirectory
100%
100%
Videos subdirectory
0%
100%
Report file present
100%
100%
Named session used
50%
100%
Wait after open
60%
100%
snapshot -i for interaction
100%
100%
snapshot without flag for content
100%
100%
Scroll via scroll command
100%
100%
Session closed
0%
100%
Report header filled
66%
100%
Initial annotated screenshot
0%
100%
72%
64%Document Issues Found in a Public Demo Site
Issue documentation and repro evidence
Video for interactive issues
0%
100%
Video started before repro
0%
100%
Video stopped after repro
0%
100%
sleep 1 between steps
0%
0%
sleep 2 before result screenshot
0%
0%
type used during recording
0%
100%
No video for static issues
0%
62%
Annotated screenshots
0%
100%
ISSUE-NNN numbering
0%
100%
Step screenshots in report
0%
87%
Issues not batched
0%
0%
fill used outside recording
100%
100%
90%
30%Full QA Pass on a Public E-Commerce Demo
Systematic exploration and wrap-up quality
Per-page errors check
0%
62%
Per-page console check
0%
62%
Per-page snapshot
0%
33%
Per-page screenshot
100%
100%
Reproducibility verified
100%
100%
5-10 issues documented
0%
100%
Severity summary accurate
100%
100%
Session closed
0%
100%
No source code reading
100%
100%
End-to-end workflows tested
100%
100%
No output files deleted
100%
100%
Final summary to user
100%
100%
- Repository
- vercel-labs/agent-browser
- Commit
fa043a4
- Evaluated
- Agent
- Claude Code
- Model
- Claude Sonnet 4.6
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.