More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Test Readers: Fresh-Eyes Feedback Before You Send

    1

    Find out how real readers will receive a draft before you send it: blind reader personas read it cold, a probe checks whether your message landed, and you get a severity-ranked report with a fix for every issue.

    Free

    13 usesSecurity scanned
    Test Readers: Fresh-Eyes Feedback Before You Send

    Test Readers: Fresh-Eyes Feedback Before You Send

    Example session with this skill installed

    Can you get fresh-reader feedback on grant-summary.md before I submit it? It goes to a regional education foundation's panel: a couple of head teachers, a school governor, the foundation's finance officer and one air-quality scientist. What I want them to take away: cheap CO2 monitors with a simple alert can keep classroom air fresh without new ventilation equipment, and we need £74,500 over 18 months to test that in 12 schools.

    • Read your context and instructions
    • Compiled the test readers

    All five readers liked the idea, and

    none of them would fund the draft as written. They got the core message (cheap monitors plus a simple alert, about £74.5k, about 12 schools). But three avoidable problems made every reader start treating the rest as a sales pitch.

    Top three issues
    1.

    The budget total is wrong. The lines add up to

    £72,500, not £74,500. I've checked the sum myself. All five readers caught it.
    2. "The pilot proved that the dashboards reduce pupil absences by 20%." Every reader stopped trusting the document at this sentence. Three classrooms over six weeks with no comparison group can't prove that. It also pushes your intended message, fresh air without new equipment, into second place.
    3. The school count changes. It's "12 primary schools" in one place and "the 10 participating schools" in another.

    The full report, with 17 ranked issues and a fix for each, is in test-readers-grant-summary-2026-10-19.md next to the draft. I haven't changed the draft. Do you want me to apply the must-fix items, or will you make the changes yourself?

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

    About this skill

    What it does

    You can't test-read your own draft: you know what each sentence was meant to say, so that is what you read. This skill gives your agent readers who don't. Point it at a paper, proposal, report or email and it:

    • Builds blind readers from five personas (Newcomer, Expert, Editor, Decision-maker, Skimmer), adapted to your real audience, and gives each one only the draft: no outline, no intent, no explanations.
    • Runs them as fresh readers in parallel when your agent supports subagents or a fresh session, so each reading is a genuine first reading.
    • Checks whether your message landed: each reader states the main message and what they think they are asked to do, set against what you intended. Drift becomes a finding even when nobody complains.
    • Ranks every issue by severity (blocks understanding, misleads, slows the reader, polish), says where it is and what the reader experienced, and pairs it with a concrete fix or question. Totals and counts the readers question are checked against the draft.
    • Keeps the tone useful: observations and questions, strengths named as concretely as problems, and your draft left untouched unless you ask.

    What a run looks like

    In our demo, five readers (governor, air-quality scientist, finance officer, head teacher, editor) read a one-page grant summary in about two minutes. They caught a budget that didn't add up, a pilot result claimed as proof, a school count that changed between paragraphs, undefined jargon and typos, and showed that the intended message had been crowded out by a claim none of them believed.

    What's included

    • SKILL.md: the full method, from scoping to the blind packet, persona prompt, comprehension probe and report format
    • examples/sample-draft.md and examples/sample-report.md: a synthetic draft and the report a real run produced
    • references/research-folder.md: the shared research/ folder convention

    Part of the Aurora Lab research pipeline

    Free, and works entirely on its own. If you keep a research/ folder, it can test-read research/paper/ and saves reports to research/reviews/. No skill calls another.

    Simulated readers know more and are kinder than real ones; the persona prompt pushes against this, but for anything high-stakes the report recommends one real reader from the audience as well. Without subagents or a fresh session, the readings happen in the same context that knows your intent, so they are less blind; the report says when this was the case. The Expert persona can be wrong about facts: treat a factual objection as something to check, not a verdict. Each persona is a separate read, so a test read uses several times the tokens of a single review, and more on long drafts. It reviews; it does not check references or rewrite the draft unless you ask. Tested on English drafts.

    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

    13 uses

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 29 days ago

    • Free with an Agensi account

    Permissions

    Read Files
    Write Files

    File Scopes

    research/paper/**
    research/claims.md
    research/reviews/**

    Reads the draft you point to (with no path and a research/ folder, it looks in research/paper/; it may read research/claims.md for your intended message). Writes one report: research/reviews/test-readers-<draft>-<date>.md, or next to the draft when there is no research/ folder. Never edits the draft unless you ask. No scripts, no network access, no installs.

    Listed29 days ago

    What's inside

    Frequently Asked Questions

    Any agent that loads SKILL.md skills, including Claude Code, Codex CLI, Cursor and Gemini CLI. It works best where the agent can run subagents or start a fresh session, so each reader meets the draft for the first time. Where it can't, the agent reads as each persona in separate passes and says so at the top of the report.

    Anything your agent can read: papers and paper sections, grant proposals, reports, cover letters, emails, and slides or PDFs if your agent can open them. Tell it who the real readers are and what you want them to take away; if you don't, it asks or states its assumptions in the report.

    No. It writes one report and leaves the draft alone. At the end it offers to apply the fixes, and only does so if you say yes.

    Each persona is a separate read, so usage grows with the number of readers and the length of the draft. In our demo, five readers on a one-page grant summary finished in about two minutes.

    No. They are a fast first test that catches most of what a careful reader would: missing asks, numbers that don't add up, unsupported claims, jargon, typos, and a message that didn't land. Real readers know less and are less patient, so for high-stakes pieces the report recommends one real reader as well.

    No. It works on its own. If you keep a research/ folder, it can test-read research/paper/ and saves reports to research/reviews/, the convention the other Aurora Lab research skills share.

    New in Writing & Content