Skip to main content

A sharp-eyed visual-audit suite for live web UIs.

Lynx audits live web UIs through real browser sessions — two coupled skills, one verdict-writer agent, 14/14 detection accuracy across two shakedowns.

Lynx is a Claude Code plugin that audits live web UIs through real browser sessions — two coupled skills produce screen-level evidence, one verdict-writer agent synthesizes a single PASS/FAIL/CONDITIONAL run-verdict.

Frozen at 14/14 detection accuracy across two synthetic shakedowns. Cycle cap fires honestly: UNFIXABLE beats a manufactured green run.

Real browsers only

Every verdict cites real screenshots, accessibility trees, and console logs captured through the agent-browser MCP. No fixtures. No DOM stubs. No mocks.

Entanglement detection

Catches multi-cycle defects: when fixing screen A unmasks a regression on screen B. WAM-class (cycle-2 reach) and synth-2-class (cycle-1 reach) both detected.

Frozen skills, honest cap

Two skills locked at 14/14 detection — modifying them requires re-running both shakedowns. Cycle cap fires honestly: UNFIXABLE beats a manufactured green run.

# Audit every screen reachable from a base URL.
/lynx:audit "http://localhost:3000"

# Single-screen audit + fix loop.
/lynx:audit "http://localhost:3000/about"
  1. 01claude plugin marketplace add krzemienski/lynx
  2. 02claude plugin install lynx@lynx
  3. 03Run /lynx:audit <base-url> to execute the full audit pipeline
Real browser audits
agent-browser MCP
Screenshot evidence
Per verdict citation
Accessibility tree capture
Entanglement detection
Multi-cycle defect classes
Frozen 14/14 detection
WAM + synth-2 shakedowns
UNFIXABLE cycle cap
Beats manufactured green
Verdict-writer agent
PASS / FAIL / CONDITIONAL
Documentation
Read the Lynx docs
lynx.withagents.dev