More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    UX Evidence: Six Questions Before You Design a Screen

    1

    This skill treats each design decision as a hypothesis.

    Free

    1 useSecurity scanned
    UX Evidence: Six Questions Before You Design a Screen

    UX Evidence: Six Questions Before You Design a Screen

    Example session with this skill installed

    We want to add an AI suggestions panel to the top of our project dashboard. Help me plan it before we design anything.

    • Read your context and instructions
    • Compiled the ux evidence

    The agent treats this as an insertion into a screen users already rely on, so it runs all six gates. It asks who opens the dashboard and how sure they are of what they want, lists three candidate pains and picks one, writes a short press release with five hard FAQs, maps assumptions and names the riskiest (that users will trust suggestions above their own data), and runs a walkthrough labeled level 0. It proposes a launch bar with numbers, such as a provisional 25% of suggestions opened within 7 days, a perception guardrail on trust, a kill condition, and a rollback. It recommends a density ladder test and flags that suggestions belong in a labeled container, not mixed into the user's own list.

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

    About this skill

    About this skill

    When the first artifact is a polished mock, the design gets judged on looks. It ships on an A/B test that looks flat, and the damage to trust shows up months later. Nobody wrote down who the screen was for, which single pain it solves, or which number would mean it should be rolled back.

    This skill treats each design decision as a hypothesis. For a new screen, flow, or module it runs six gates before UI code: the user and the evidence that they have the problem; one pain chosen from three ranked by frequency and severity; a three-sentence press release with five hard FAQs; an assumption map tagged by value, usability, feasibility, and viability risk; a cognitive walkthrough; and a launch bar with a success metric, a perception guardrail, a kill condition, and a rollback. A new component in an existing flow gets two gates, and a copy tweak gets none. It also picks the research method that answers a given question, from ten method cards, and checks a design against 23 principles on density, trust, labels, and attention.

    The proof

    There is no measurement for this skill, and the listing claims none. Its checks are mechanical. Thresholds must be numbers: "TBD" and "X%" are not allowed, and with no baseline it proposes a starting number labeled provisional, saying what will revise it. A walkthrough the agent runs alone is labeled level-0 evidence and must be confirmed with one real person. Every sample size and threshold marked as a starting assumption is labeled that way so it is never quoted as a finding. The README credits the published methods it builds on, including the Working Backwards press release and FAQ, the cognitive walkthrough, story mapping, assumption mapping, Marty Cagan's four risks, and Nielsen Norman Group guidance.

    What you get

    SKILL.md with the gates, method picker, six design checks, and rules for forms colleagues fill in; reference/principles.md (23 principles); reference/methods.md (10 method cards with setup, sample size, what to measure, when not to use it, and a solo version); reference/measure.md (question bank, tracker blueprint, experiment guardrails, metric criteria); and templates for a research plan, a design review, and a findings report.

    How to use

    1. Add the skill folder to your agent's skills directory (for Claude Code, ~/.claude/skills/ux-evidence/).
    2. Before designing a new screen or flow, ask the agent to plan it, or ask which research method answers your question.
    3. Put the gate answers at the top of your spec, then hand visual design to a design skill or designer.

    Source and license

    Free to use and MIT licensed. The full source, install steps, and issue tracker are on GitHub at https://github.com/robertmnyborg/ux-evidence.

    Plans and judges UX work; it does not recruit participants, run studies, or produce visual designs. Agent-run walkthroughs count only as level-0 evidence. Sample sizes and thresholds marked as starting assumptions are defaults, not benchmarks. Hand-offs to other skills only happen if those skills are installed. No measured results yet.

    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

    1 use

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 8 days ago

    • Free with an Agensi account

    Requires

    Any agent that loads SKILL.md files
    Optional companion skills: impeccable for visual design, plus trustworthy-insights, causal-claim, and job-card

    Permissions

    No special permissions declared or detected

    Listed8 days ago

    What's inside

    Frequently Asked Questions

    Agents asked for a screen produce a screen. They rarely ask who it is for, which one pain it solves, or what number would kill it. This skill requires those answers first and scales the process down to nothing for small tweaks.

    Yes. It is markdown only, with no scripts or hooks, so any agent that reads SKILL.md can use it.

    No. It decides what to build and how you will know it works. It hands visual craft to a design skill such as impeccable, or to you.

    Yes. Each method card includes a solo version for one builder plus an agent, such as an unmoderated expectation test with 10 people per variant.

    On GitHub at https://github.com/robertmnyborg/ux-evidence, under the MIT license. Every file in this package is there, plus the README and install steps.