Reference
Field Suit
A runnable reference workflow where an AI can prepare an impact report but cannot turn that capability into authority: approvals bind to exact content, destinations default closed, and source instructions are treated as data.
Who it is for — and why
Forward-deployed builders designing AI-assisted work that must cross from prototype to accountable handoff without hiding the approval and egress boundaries.
Proof you can inspect
- Thirty-three offline synthetic cases exercise approval binding, prompt-injection quarantine, destination allowlists, cost ceilings, and fail-closed behavior
- The reference has zero runtime dependencies and needs no credentials or network connection
- Read-only source adapters expose no write method, so the model cannot acquire mutation through configuration
- A deterministic public-export inventory fails on private paths, unexpected top-level content, or changed file membership
Limits
- Not a validated client platform
- Does not establish regulatory compliance
- The public repository is the exported reference only — the working platform stays private