Demo on fictional data
Isolation leak test
This is a demo on fictional data that shows a leak test between two made-up clients: twelve queries run against Client A's workspace, and any result from Client B's documents fails the test.
Requests since you opened this page: 0
Works offline: not installed
The browser policy on this page forbids network requests (connect-src 'none'). The count is here so you can see it hold. How the demos work.
Two client workspaces
Each made-up client has its own documents and its own keyword index, held separately in this page's memory. Client B's documents include a planted marker, CANARY-B-7731, that must never be found from Client A.
Client A: Sample Landscaping Co. (fictional)
Client B: Sample Dental Office (fictional)
Search Client A's workspace
The leak test
Twelve queries run against Client A's workspace. Any result that comes from Client B's documents fails the test, and one failure blocks the release.
| # | Query from Client A | Hits from A | Hits from B | Result | Matched text |
|---|
What this shows, and what it does not
What this shows
- Separate in-memory indexes per client, the shape of FoxGrade rule 3.
- A planted marker in Client B that must never be found from Client A.
- What a failing run looks like when isolation is broken on purpose.
What it does not show
- The index is a simple keyword index in your browser. The point is the test, not search quality.
- Passing here says nothing about any real system.
- Both clients are made up.
How it would run for you
This is the shape of a FoxGrade: the agent drafts, a named person approves, the log stays with you.
Related services: FoxGrade, Production Readiness Review, Architecture Sprint.