UX Health Snapshot
A compact read on the overall state of the interface, including the clearest usability risks and evidence limits.
Packet page
Start Here
A quick orientation for UX people: what this is, why it exists, and where it fits in a real UX workflow.
What It Is
UX Heuristics Compass is an audit assistant for visible user experience evidence. It can review a website page, app screen, prototype screenshot, small page capture, or bounded task-flow note and turn that evidence into a scored usability readout.
It is meant for the moment before a design review, handoff, portfolio critique, stakeholder conversation, or early QA pass, when you want a sharper read than generic AI commentary but you do not yet have time for a full research cycle.
What It Produces
Heuristic Scorecard
Ratings organized around Nielsen Norman Group's core heuristics plus optional accessibility, inclusion, journey, and UX-writing profiles.
Prioritized Fix List
A ranked set of issues with observed evidence, severity, recommendation, confidence, and suggested next validation.
Where It Fits
This does not replace usability testing with real people. It helps you catch obvious or likely UX problems earlier, prepare better review conversations, and decide what to validate next.
Packet page
How It Works
The tool keeps the UX review grounded: first in visible evidence, then in heuristic structure, then in human calibration where the AI needs context.
The Audit Flow
- Start with bounded evidence: a screenshot, page capture, small interface sample, or clearly named task flow.
- Add just enough context: audience, task, platform, and whether optional profiles such as accessibility or UX writing should be included.
- Run a quick audit or fuller checklist audit to produce ratings, issues, evidence limits, and recommendations.
- Use human-in-the-loop questions when the audit needs mission context, target-user nuance, or confidence calibration.
- Turn the output into a stakeholder-ready report only after findings have evidence references, severity, recommendations, and confidence.
The Human Part Still Matters
Context changes severity
A loading delay, hidden exit, or lost progress path can be minor in one product and critical in another, depending on the audience and task.
Evidence boundaries stay visible
The report names what was and was not captured so readers do not mistake a bounded heuristic audit for a full research study.
Next validation is part of the output
The tool should point to the next real check: a usability test, accessibility scan, engineering verification, or follow-up research question.
Audit Modes At A Glance
| Mode | Use it when | Typical output |
|---|---|---|
| Quick audit | You need a fast read on a screenshot or small page capture. | UX Health Snapshot, top findings, one concrete next action. |
| Full checklist audit | You want scored findings across the active heuristics. | Per-heuristic scorecard, issue ledger, evidence limits, fix order. |
| Human-loop audit | The interface context affects severity or report readiness. | Calibration questions, revised priorities, report-ready payload guidance. |
| Multi-perspective audit | You want several expert lenses without pretending there was a real user study. | Evaluator disagreements, shared priorities, evidence that would change the rating. |
Packet page
Try It Yourself
You can try the prompt-only version today while the downloadable packages remain locked for beta verification.
Start With One Screen
Pick one bounded artifact: a screenshot, a prototype frame, a public page capture, or a short task-flow description. Avoid asking for a full-site audit on the first pass.
The first useful result should identify the top usability risk, what evidence supports it, what the audit cannot know yet, and the smallest worthwhile fix.
Quick Audit Prompt
TEXT
Use UX Heuristics Compass for a quick audit.
Review this interface evidence:
"""
[Paste screenshot notes, page text, URL capture notes, or a short task-flow description here.]
"""
Audience:
[Who is this for?]
Task:
[What should the user be able to do?]
Focus on visible usability issues, likely user confusion, severity, evidence limits, and the top fix to make first.
Return a compact UX Health Snapshot, the top 5 findings, and one concrete next action.Good First Inputs
- One screen or one task flow, not a whole product.
- A clear user type or audience.
- The task the interface is supposed to help with.
- Desktop, mobile, or both.
- Any known constraints such as low bandwidth, accessibility needs, child users, high-stakes completion, or first-time use.
Current Beta Status
The public landing page is live, but package downloads are intentionally disabled until plugin, MCPB, skill, source, and checksum assets are verified. Use the prompt path first.
Next Step
Packet page
Example Output
The Oppia.org example shows the kind of public-safe report UXHC can produce when it has bounded page evidence and human calibration.
Oppia.org Scorecard Example
What The Example Demonstrates
Prioritized severity
The report does not just list issues; it elevates progress loss as the number-one fix because it breaks the learning loop.
Mission-aware calibration
Severity is interpreted through the product context: free education for under-resourced learners changes how loading, progress, and navigation problems should be weighed.
Evidence limits
The report names what was captured and what still needs validation, including accessibility testing, mobile checks, and engineering confirmation.
Oppia Audit Snapshot
| Signal | Result |
|---|---|
| Overall grade | B- / 68.5%, below average with multiple issues. |
| Top finding | Progress is lost for all users, including logged-in accounts. |
| High-priority themes | Progress persistence, lesson loading feedback, visible exit path, screen-reader loading state. |
| Important caveat | The audit was bounded to selected pages and is not a full WCAG compliance test. |
Open The Example
- View the Oppia UX audit report: Public example report route on the UX Heuristics Compass site.
- Open UX Heuristics Compass: Landing page with prompts, beta status, install expectations, and FAQ.
Packet page
Technical Notes
This section explains the AI and MCP layer in practical terms without requiring readers to know protocol details.
What MCP Means Here
MCP stands for Model Context Protocol. In plain language, it is a way for an AI assistant to connect to a real local tool instead of only replying with freeform text.
For UXHC, that means the host AI can use structured local helpers for source preparation, heuristic lists, audit scoring, validation, report generation, preferences, and multi-perspective review. The user still decides what evidence to provide and what to trust.
Division Of Labor
The host AI
Interprets screenshots, asks questions, explains findings, and adapts the conversation to the reviewer.
UXHC
Owns the rubric data, tool contracts, scoring structure, report-readiness checks, and local report artifacts.
The human reviewer
Provides context, checks whether severity makes sense, validates findings with real users or implementation evidence, and decides what ships.
Important Boundary
UXHC is a heuristic review assistant, not a full UX research platform, crawler, analytics system, or autonomous browser agent. Its beta scope is intentionally bounded so outputs stay reviewable and useful.
Current Package Status
| Surface | Status |
|---|---|
| Prompt-only use | Available through the public page prompts. |
| Claude Cowork plugin | Planned, pending package build and clean scan. |
| Claude Desktop extension / MCPB | Planned, pending manifest and tool-surface verification. |
| Skill and source packages | Planned, pending review and checksums. |
Packet page
Feedback
The useful feedback is not whether the page sounds impressive. The useful feedback is whether a UX practitioner would trust it, try it, and know where it belongs.
Questions For UX Practitioners
- Does the Oppia example feel like a useful heuristic audit output, or does it overstate what the evidence can support?
- Which part would you use first: quick prompt, full scorecard, human-loop calibration, or report generation?
- Where does the explanation become too technical for a UX audience?
- What would need to be clearer before you used this in a design review or stakeholder handoff?
- What is the first output example this should add next: mobile app screen, onboarding flow, SaaS dashboard, nonprofit site, or prototype frame?
What To Send Back
One strength
The part that made the tool feel useful, credible, or worth trying.
One friction point
The part that felt confusing, too technical, too generic, or not grounded enough.
One next example
A type of interface or UX situation that would make the tool easier to judge.
Best Next Step
Open the UXHC page, skim the Oppia example, and try one quick audit prompt on a screenshot or flow you already know well. That will show whether the tool helps you think more clearly or just adds ceremony.
Next Step
Next Step
Open UX Heuristics CompassView the Oppia Example ReportTry a Quick Audit Prompt