browser-verifier — Claude Code subagent: what it does, how to install
browser-verifier — a subagent that verifies client-side vulnerabilities in a real browser. Takes a curl-based finding and determines whether it executes in the browser.
no license ★ 1,298 updated 2026-06-12
.claude/agents/browser-verifier.mdWhat browser-verifier does
The subagent runs XSS, DOM, postMessage and prototype pollution checks in a live browser via camofox (headless Firefox). The process: detect Cloudflare and WAF, spawn a tab, inject a detector (DOM marker, fetch callback or OOB exfiltration via dns-prefetch), navigate to the payload URL, snapshot the page and check whether it fired. If it did not fire, diagnose the reason: CSP blocked it, framework escaped output, WAF challenged the browser, or the context is non-executing.
Returns a verdict: BROWSER_CONFIRMED (executed), BROWSER_REJECTED (reflected in the response but the browser did not run it), or BROWSER_PARTIAL (executed but CSP limits exfiltration). Reflection in an HTTP response does not prove execution: React and Angular escape output automatically, CSP blocks eval and inline scripts, and some WAFs pass curl but challenge browsers. The subagent is dispatched automatically by /hunt and /validate for client-side vulnerability classes.
Requires a running camofox instance and access to localhost:9377. The $CLAUDE_PROJECT_DIR/tools/camofox_ctl.sh script starts the browser if it is not running.
What it helps with
- verify whether an XSS payload executes in a real browser
- distinguish reflection in the response from actual DOM execution
- diagnose why CSP or a framework blocks the payload
- capture a screenshot of the fired vulnerability as evidence
- bypass WAF using a headless browser with a stealth profile
Example prompts for the agent
- Verify finding
xss-reflected-001.jsonin the browser and return a verdict - Run browser-verifier on
findings/dom-xss-checkout.md - Diagnose why the payload in
xss-profile.jsondid not fire in Firefox
How to install browser-verifier
Manually: put the files in place
.claude/agents/browser-verifier.mdFor all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention
Manually: put the files in place
.codex/agents/browser-verifier.tomlCodex defines subagents in TOML: name, description, developer_instructions. Move the Markdown text into developer_instructions by hand. For all projects use ~/.codex/agents/
Manually: put the files in place
.cursor/agents/browser-verifier.mdCursor also reads .claude/agents/
Step by step, with commands for each platform: Claude Code subagents: how to create one, call it and keep context clean
Questions about browser-verifier
How do I install browser-verifier in Claude Code?
Place the .claude/agents/browser-verifier.md file in the project root. The subagent is dispatched automatically by /hunt and /validate for client-side vulnerabilities.
What does browser-verifier do?
The subagent verifies client-side vulnerabilities (XSS, DOM, postMessage, prototype pollution) in a real browser via camofox, snapshots the DOM and determines whether the payload executed. It distinguishes reflection in the HTTP response from actual execution in the browser.
Why verify in a browser if the payload already reflects in curl?
Reflection in the response does not prove execution. CSP can block inline scripts, React and Angular escape output automatically, a WAF may pass curl but challenge the browser, and the context may be non-executing. browser-verifier checks actual execution and diagnoses rejection causes.