Motion & effects / Reference-to-prompt library
Kage
Turn a real interface reference into an explicit starting brief.
Understand the resource
What it is.
Kage organizes real product interfaces, components and page types and offers prompts for coding tools. The useful step is translating an observed layout into requirements that can be reviewed against your own product.
When it is useful
An agent needs a concrete composition target for one screen or section.
A proposed application
Put it to work.
This example illustrates a possible use. The proposed workflow and expected result are separate from the bounded checks recorded below.
Try this
Pick a pricing reference, describe the hierarchy you want, and replace the original offer and identity before generation.
Result to aim for
A specific layout brief rather than a vague aesthetic request.
Before you start
What you need.
- A chosen reference and a project brief with owned content.
The practical boundaries
What to watch for.
- Recreation prompts do not transfer rights to the reference design, copy or assets.
- Generated output still needs semantic, responsive and accessibility review.
Review record
What we could establish.
- Public primary documentation and the bounded account-free checks below were reviewed. The public search field accepted dashboard, but the bounded search probe did not establish a changed result set. Reference browsing was reachable; search effectiveness remains unconfirmed.
The source-review notes describe documentation and identity. The account-free evaluation below records exactly which public interaction or local fixture was checked, what happened, and what remains untested.
Check current access, terms, licensing and data handling before using a resource with your own material.
Runtime and security evaluation
What the evaluation covers.
The public search field accepted dashboard, but the bounded search probe did not establish a changed result set. Reference browsing was reachable; search effectiveness remains unconfirmed.
Methods: Public HTTP response · Public browser sample. A completed check covers the stated sample; it is not a safety certification.
- Target
- https://kage.design/
- Version or capture
- Unversioned public page checked 2026-10-19
- Integrity scope
- sha256:823e60912a776e2e7692a90b95ee8c3eb86c017399f5051c58cca0b3ab63a024
Test conditions
- Chromium 155.0.8059.82 on Windows; fresh temporary profile and fresh account-free contexts.
- Public HTTPS GET/HEAD requests only; telemetry/auth routes and WebSockets blocked. No account storage or credential environment variables.
Individual checks and observations
Public document and DOM
passPublic browser sample: Fresh Chromium GET, then inspect rendered controls after hydration delay
Observed: HTTP 200; 15 captured button labels; 2 inputs; 1 canvases.
Reference search
failPublic browser sample: Type dashboard into the public search field without submitting a site
Observed: Visible links 195 -> 195 after the synthetic GET search. Search-result relevance not independently scored; submission fields untouched.
Security findings and boundaries
Passive inspection of the public response headers, requested hosts and the named browser sample. No active exploitation, authenticated permissions test or complete security audit.
Observed · caution
Search acceptance is weaker evidence than search results
The query was retained while the global visible-link count stayed195. This may reflect the page structure or checker scope; it is not proof that search is broken.
Practical caveats
- Inspect the reference manually and write an original brief with owned content; do not assume a search result grants reuse rights.
- Pin and recheck the selected version or sample before adoption; the observations apply only to the scope and date shown.
What remains untested
- No confirmed search-result relevance, prompt generation, model execution or site submission.
- Fresh-browser restrictions can affect external assets and hydration. Initial failed selectors and limited probes remain in the evidence; they are not blanket product failures.
Security limits
- No authenticated workspace, private content, production backend permissions or data-retention verification.
- No confirmed search-result relevance, prompt generation, model execution or site submission.
Evaluation sources
Follow the evidence
Sources and further reading.
Keep exploring