External qualification workbench

Test the ugly case without sending it anywhere.

Select one failed engagement. This page hashes the surviving record locally, exposes privacy holds and missing evidence, and shows the exact analysis floor Tier Desk would have to run before making any claim.

No account, upload, API call, model call, client contact, or source-system writeback.

Preview one engagement

Choose a folder, selected files, or the synthetic failed case.

Drop the surviving engagement record here

Proposal, messages, tickets, estimates, invoices, architecture notes, tests, repository metadata, scope changes, and final outcome. Contents remain in this browser tab.

No files selected.

Preferred custody

RedCat keeps raw evidence local. Export only the manifest or pseudonymized preview it chooses to release.

Automatic refusal

Credential-shaped evidence produces HOLD_REDACTION_REQUIRED.

Pass condition

Architects can trace the diagnosis, implement the packages, and agree the counterfactual would have reduced unpaid work.

Case Zero intake preview

This is the evidence and execution floor, not a retrospective conclusion.

Files0

Bytes0

Sensitive holds0

Corpus digest-

Evidence coverage

How the floor would run

Job	Object	Preview state	Evidence	Output

Source register preview

Path	Class	Bytes	SHA-256	Signals

A real Case Zero result must bind every material statement to evidence, preserve missing and contradictory facts, compare actual and counterfactual economics, and survive RedCat architect challenge.

Runner command

python casezero.py intake ^ --input "D:\RedCat\CaseZero\raw" ^ --output "D:\RedCat\CaseZero\intake" ^ --case-id "REDCAT-CZ-001" ^ --custody-mode "redcat_local" ^ --as-of "2026-09-01T12:00:00Z"