BB

    Browser Bug Validator Companion

    1

    Guidance for the open-source project's recorded trace demo and local fixture evaluations. External runtime required.

    Free

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more
    BB

    Browser Bug Validator Companion

    Example session with this skill installed

    Using the pinned repository, check the documented software and HTTP controls and attempt the scripted browser control. Report what actually ran.

    • Read your context and instructions
    • Compiled the browser bug validator

    Verified during companion preparation against commit 6be965fa69c0b884613ddd72249c02524efead93: 185 Python unit tests, 13 viewer tests, 5 calibration-tool tests, 17 catalog definitions, and 11 HTTP goldens passed. The scripted browser attempt retained all 11 cases: two pre-browser refusal controls passed, but Chromium failed to launch for the other nine, so the overall browser gate failed and no live-browser workflow passed locally. No local Ollama server was running, so no model trial ran. The existing upstream CI for that exact commit was separately verified successful. These are software/integration facts, not a new model-reliability result.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    A free, guidance-only companion to the Browser Bug Validator GitHub project. Learn how to inspect its offline historical trace viewer, check runtime prerequisites, run software and scripted browser controls, and report local-model trials without confusing test doubles with real inference.

    The package includes SKILL.md, a local runbook, provenance notes, and the retained MIT license. It does not include or host the repository, agent runtime, browser, model, or inference service. Loading the skill does not install dependencies or run a test. Commands are pinned to public commit 6be965fa69c0b884613ddd72249c02524efead93.

    This is an unreliable research prototype for authorized disposable fictional loopback fixtures. The recorded general-loop baseline remains 5/16 full browser-validation trials (11/22 including six refusal/access controls). The historical trace viewer and passing scripted controls are not new model-reliability evidence. Defect Reproduction Check remains a separate skill.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse All Claude Skills.

    Guidance only, no hosted execution or bundled runtime. Evaluation runner does not support Windows. No arbitrary or production website support, credentials, paid inference, or Bedrock adapter. Model reliability remains unproven. Package validation and software/HTTP controls passed; this preparation environment could not launch Chromium and had no Ollama server, so no fresh live-browser or model pass is claimed.

    How to use

    1. 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. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 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

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 8 days ago

    • Free with an Agensi account

    Accounts
    GitHub0
    Software
    Python 3.10+Node.jsPlaywright 1.48+ChromiumGoogle ChromePython 3.12Playwright 1.65.0OllamaGit

    Separate Browser Bug Validator repository checkout. documented revision 6be965fa69c0b884613ddd72249c02524efead93. Recorded viewer only: downloaded repository and a modern browser. Local process and loopback-server access. browser sandbox must remain enabled.

    Permissions

    Allowed Hosts

    github.com
    127.0.0.1

    File Scopes

    Separate
    repository
    checkout
    Machine-local
    browser/model
    configuration
    Fresh
    local
    evaluation
    run
    directory

    No automatic installation, dependency/model download, production access, payment, external upload, or profile change. Local runs can contain supplied report/KB/page text. Sharing artifacts requires separate authorization.

    Listed8 days ago

    What's inside

    Frequently Asked Questions

    No. This package delivers instructions. Runtime, browser, repository checkout, and local model execution are separate.

    Yes. The downloaded repository's static trace viewer opens in a modern browser. It shows selected historical text evidence, not fresh validation.

    No. It is a controlled-fixture research prototype with a recorded 5/16 full-browser-trial model baseline. New sites require separate reviewed capabilities, fixtures, and tests.

    No. It is a separate companion to the Browser Bug Validator GitHub project.

    Free