More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Universal Verification

    1

    Forces agents to prove every “done” claim with real unit, API, browser, visual, accessibility and performance evidence

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Universal Verification

    Universal Verification

    Example session with this skill installed

    I just finished the new pricing table on the landing page. Verify it is actually done and ready to ship.

    • Read your context and instructions
    • Compiled the universal verification

    Verification Report

    • Unit: PASS (npm test -- pricing → 12 passed, 0 failed)
    • API: SKIPPED (no new endpoints)
    • Browser: PASS (https://staging.example.com/pricing — table renders 3 tiers, annual toggle works, console clean)
    • Visual: PASS (matches Figma spacing + colors on 1440px and 375px)
    • Accessibility: PASS (axe: 0 critical/serious; keyboard focus order correct; all prices labelled)
    • Performance: PASS (LCP 1.4 s, no bundle growth > 2 kb)

    Status: VERIFIED
    Unproven: none

    Next steps

    • Merge the PR
    • Tag the release
    • Optional: add visual regression snapshot to CI

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

    About this skill

    Universal Verification is the proof gate every serious agent workflow needs.

    Most agents love to say “it works”, “tests pass”, or “ready to ship”. This skill refuses those statements until executable evidence is produced across six layers:

    1. Unit / Logic – full relevant test suite (or write the missing tests first)
    2. API / Contract – real calls with status, body and latency
    3. Browser / End-to-End – actual user flow + final URL + visible state
    4. Visual – layout, design tokens, regression check
    5. Accessibility – keyboard, labels, contrast, critical issues
    6. Performance – latency, bundle size, regressions

    The agent must show raw evidence, not summaries. Status is only VERIFIED, PARTIAL or FAILED. Claims without proof are rejected on the spot.

    Perfect for:

    • Teams that ship with AI agents and need a hard quality gate
    • Code-review and PR workflows that demand proof
    • Any “is it done?” moment before release

    Install once. Every future “feature complete” claim becomes measurable.

    Works only with agents that support the open SKILL.md standard and can run tools (browser, terminal, network).
    Does not invent evidence — if a tool or environment is missing, that layer is marked UNPROVEN / SKIPPED.
    Cannot force verification if the user explicitly says “skip verification”.
    Best results when the agent has access to a real project, tests, and browser tooling.
    No external API keys or paid services required.

    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 any payment.

    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 1 month ago

    Permissions

    No special permissions declared or detected

    Listed1 month ago

    What's inside

    Frequently Asked Questions

    This skill prevents 'hallucinated completion' by requiring raw command outputs, logs, and status codes. Instead of an agent simply claiming a task is done, it must provide verifiable evidence from unit tests, API responses, and browser audits.

    Universal Verification is designed to detect your existing stack (like Jest, Pytest, or Playwright) and run the appropriate commands automatically. It requires the agent to have access to a terminal environment where these tools are installed and executable.

    Yes, it includes visual and accessibility audits that use browser automation to check for layout breakage, keyboard navigability, and color contrast. It ensures that UI changes meet modern standards rather than just 'looking right' to the agent.

    The skill is compatible with most advanced AI agents that have file system and terminal access, such as Claude Dev, AutoGPT, or custom LangChain implementations. It functions as a supervisory layer that intercepts the 'task complete' signal until evidence is provided.

    If your environment lacks specific credentials or tools (e.g., a missing API key for a production-only test), the skill will flag these layers as 'unverified' in the final report rather than failing the entire process. This allows you to see exactly which parts of the code remain unproven.

    Those layers are marked SKIPPED / UNPROVEN and the overall status becomes PARTIAL. The skill never invents evidence. It only reports what it can actually run.

    No. If you explicitly say “skip verification” or “just claim it”, the skill steps aside. Otherwise it enforces the proof requirement.

    It works with any agent that supports the open SKILL.md standard and can use tools (browser, shell, network). It has no hard dependency on a specific model or platform.

    No. It only activates when the agent is about to claim completion, or when you explicitly ask to verify / prove / check the work. Routine coding stays fast.

    The skill requires the agent to write and run the missing tests first (partial TDD). It will not accept “it should work” without green tests covering the new behavior.

    Only if the change is already implemented in a running page the agent can open. Pure design files without a live implementation cannot be fully verified.

    By default it stops on the first hard failure so you can fix the issue. You can tell it to continue collecting evidence if you prefer a full report.

    No. It is a hard gate. Claims without executable evidence are rejected on the spot. The agent must show real command output, status codes, screenshots, or accessibility results.

    It is designed for agent conversations. You can still use the same verification mindset in CI, but the skill itself runs inside the agent session.

    New in Testing & Debugging