More screenshots
Works with the AI Agents you already use
Browser Verified Skill
Drive a real browser with evidence: every action passes a policy gate, every run compiles into a replayable Playwright test.
$29
browser-verified-skill
Example session with this skill installed
Log in to my dashboard, confirm the 'Active Projects' panel shows, and save a replayable Playwright test for CI.
- Read your context and instructions
- Compiled the browser-verified-
session started
snapshot read: 14 elements, refs assigned
filled 'role=textbox name="Email"' -> ok
filled 'role=textbox name="Password"' -> ok
click 'role=button name="Sign in"' -> exit 4 (needs approval)
approval id: apr_7f3a — a human approved it out of band
click 'role=button name="Sign in"' --approve <token> -> ok
verify { text_visible: 'Active Projects' } -> VERIFIED: every condition passed
compiled -> flows/dashboard.spec.ts (plain Playwright, no `bv` import)
The run is auditable, the success is proven, and the test replays free.
Connects securely to your tools. The creator never sees your data.
About this skill
Drive a real browser with evidence, not guesswork.
bv is a browser driver for AI agents where every action is checked before it runs and every run leaves an auditable artifact behind.
What makes it different from any other browser skill:
- A policy gate on every action. A plan declares the goal, the allowed origins and the allowed action classes. Anything outside it is refused. Exit 3 means stop — the action is outside the plan, and retrying it will not help.
- Prompt-injection detection. If a page contains hidden instructions, the run reports an injection finding (exit 5). The page is treated as data, never as instructions.
- Human approval for consequential actions. Exit 4 means a human must approve. The refusal names an approval id, and the one-time token is bound to that exact action.
- Verification instead of hope. Do not report success from an action's exit code.
bv verifychecks a real condition, and an unverified run is reported asunverified— never as done. - It compiles into a plain Playwright script. The run ends as a
.spec.tsthat imports nothing frombv, so it runs in any Playwright project and in CI, for free. Replay a recorded run with no LLM cost.
Honest about its limits. The injection guard is a speed bump, not a wall. A CAPTCHA or an impossible login means hand off to a human — there is no solver. The browser is Chromium; there is no stealth, no fingerprint spoofing and no proxy rotation, by design. It does not defeat authentication, paywalls, WAFs or rate limits, and it refuses to.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Security.
The injection guard catches known patterns, so a novel one may pass — it is a speed bump, not a wall. The policy gate is conservative by design, so some legitimate flows need a human approval an automated agent cannot give itself. Chromium only. No CAPTCHA solving, no stealth, no proxy rotation. It does not defeat authentication, paywalls, WAFs or rate limits.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 4 days ago
- Software
Node.js 18+npmChromium (auto-installed)
- Optional
- A human for approvals and CAPTCHAs
The skill installs the CLI itself with `npm install -g browser-verified-skill` if it is missing. The browser is downloaded on first use.
Permissions
Allowed Hosts