Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Web Research To Sourced Brief

    1

    One markdown research brief where every claim links to a page the skill actually opened, with the sentence it came from, disagreements shown and gaps listed

    $5

    /mo

    Subscription · cancel anytime

    Secure checkout via Stripe

    • Always the latest version
    • Delivered live to your agent via MCP
    • Cancel anytime, access ends at period end
    Security scanned
    Web Research To Sourced Brief

    Web Research To Sourced Brief

    Example session with this skill installed

    Research how browsers are doing on passkey support right now and write it up with sources — I have to present this to our security team on Thursday and they will check the links.

    • Read your context and instructions
    • Compiled the web research to

    research/browser-passkey-support.md

    Summary

    • Every major browser engine creates and asserts passkeys; the Web Authentication API that passkeys are built on is documented as available across browsers since September 2021 [S7]
    • The version floors most often quoted are Safari 16+, Chrome 108+, Edge 108+ and Firefox 122+, but that figure comes from a vendor, not from a browser maker [S11]
    • Firefox closed the last well-known browser feature gap on 2026-06-16, when Firefox 152 shipped WebAuthn Related Origin Requests [S8]
    • Where the credential is stored, not whether the browser supports passkeys, is what varies: Chrome writes to Google Password Manager by default on every desktop OS [S3], and third-party providers on Android require Android 14 or later [S6]
    • Two providers hold most of the world's passkeys - iCloud Keychain at an estimated 44-69 percent and Google Password Manager at 21-33 percent [S15]
    • Desktop is the weak surface: identifier-first passkey login on Windows 10 and 11 is reported at 45-60 percent success, against first-time enrollment of 25-39 percent on Windows web versus up to 83 percent on iOS web [S15]
    • Passkey portability between providers is still specification work: FIDO publishes the credential exchange specifications [S5] but no captured page states that the transfer protocol has been finalized [S10] [S12]
    • A large enterprise directory still has concrete browser holes - Microsoft Entra ID's own matrix marks Firefox on Android unsupported, and passkey sign-in via Microsoft Authenticator unsupported in Firefox on Linux [S4]

    Settled: All four major engines create, assert and autofill passkeys, and the last advanced-capability gap closed when Firefox 152 shipped Related Origin Requests on 2026-06-16; what differs between browsers is where the credential is stored and whether a third-party manager may hold it.

    Disputed: Whether the remaining problems are solved — HYPR's CEO, quoted in ID Tech, calls account recovery "a solved problem" and provider lock-in a fear "dressed up as a security concern", while MojoAuth's matrix calls a single device-bound passkey "a lockout waiting to happen" and an engineering practitioner blog calls fragmentation the thing that "kills adoption"; MojoAuth also contradicts itself on Firefox, listing it at 122+ with cross-device support in July 2026 and at 109+ with "inconsistent" cross-device support in March 2026.

    Largest gap: No Apple- or WebKit-published page was captured, so every Safari and iCloud Keychain claim here is second-hand from vendors and reporting.

    $5/mo

    Subscription · always the latest version

    • Secure checkout via Stripe
    • Cancel anytime
    • Delivered live via MCP

    About this skill

    Fifteen tabs open, and half the claims have lost track of where they came from. This skill opens every page it cites, in a headless browser when it can, then writes one markdown file where each claim carries a live URL and the verbatim sentence behind it. Sources are typed - primary, reporting, vendor - and four articles resting on one announcement count as one. Where sources disagree, both sides are shown, not averaged. What it could not find on a page is listed as a gap, not written as fact.

    What it does

    Give it a subject and it comes back with one markdown file you can hand to someone else: a summary, findings grouped by sub-question, a disagreements table, a numbered source table with full links and published dates, and a gap list.

    Try it first

    Say "research the current state of WebGPU browser support and write it up with sources". You should get back a path to research/webgpu-browser-support.md, plus three lines in chat saying what the sources settle, what they disagree on, and the biggest gap. Open the file: every bullet ends in a source number, every finding carries a quoted sentence, and the source table has full https:// links you can click and check. If the Gaps section says the pages were captured as text without a browser, Playwright is not installed: the brief is still sourced, but pages that build their content after load may come back thin. Install it for the full capture.

    What it needs

    A web search tool to find candidate URLs, a web fetch tool, and Write to save the brief. Recommended: Python 3.8 or newer with the playwright package and Chromium installed through it (pip install playwright, then playwright install chromium - roughly 150 MB, once), plus Bash to run the script - that gives a real browser capture, which reads pages a plain fetch cannot. Without Playwright the skill fetches each page as text instead, quotes only from what came back, and records in Gaps that no browser was used. No API keys, no MCP servers, no accounts.

    It does not log in and does not work around paywalls - a page that wants an account is recorded as a gap, and on subjects where the good sources sit behind paywalls the brief will be thin. Some sites build their content in the browser after the page arrives, or route it client-side; caniuse.com is one, and a direct hit there returns the site shell rather than the feature table. The capture flags those as having little readable text and the brief lists them as gaps, so you get a visible hole rather than a confident sentence with a link that does not support it. It is only as good as the candidate URLs it starts from, so an obscure subject with little written about it produces a short brief with a long gap list, which is the honest answer but not always the wanted one. It will not give you a live number - a price, a score, a count that moves hourly - because a brief is dated the moment it is written. It does not decide anything for you: when two sources disagree it shows both and stops, and the call stays yours. Sites that block automated browsers outright will fail to capture, and those failures are reported rather than quietly skipped. Without Playwright, client-rendered pages lose more of their text, and those also land in Gaps rather than being cited thin.

    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

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 28 days ago

    Requires

    Python

    Permissions

    Terminal / Shell
    Browser
    Network Access
    Read Files
    Write Files
    Listed28 days ago

    What's inside

    Frequently Asked Questions

    One markdown research brief where every claim links to a page the skill actually opened, with the sentence it came from, disagreements shown and gaps listed

    This skill uses the open SKILL.md standard. Use it through an MCP-compatible agent connected to Agensi, and check the compatibility information on this page for confirmed agents and required capabilities.

    Connect your AI agent to Agensi through MCP, then ask it to use Web Research To Sourced Brief on your task. Your agent loads the latest approved version each time. Paid skills require a purchase or active subscription. Skill files are not available for download; your licence does not allow you to save, copy or export any part of a skill.

    New in Research & Analysis