scope-path-control
Scope & path control
Every change is bounded before it starts: approved intent, allowed paths, and protected files and secrets an agent may never touch.
Connects to: Protected execution, Rollback & human authority
assurance
Assurance
NEO turns AI-generated software change into a controlled, inspectable decision process. It does not certify an absolute result; it produces re-auditable grounds for a keep, remediate, reject, or rollback decision.
Agent confidence never constitutes evidence. Acceptance requires reproducible artifacts, independent judgment, and human authority.
The six control surfaces
protected-execution
Protected execution
Work runs inside enforced boundaries — write mediation and snapshots before any destructive step, nothing beyond the approved surface.
Connects to: Scope & path control, Evidence custody
frozen-acceptance
Frozen acceptance criteria
Acceptance tests are authored and frozen before implementation, so the bar can't move to fit the result; judges are checked for vacuity.
Connects to: Independent audit, Evidence custody
independent-audit
Independent audit
A cold, context-isolated channel re-runs the evidence — not the builder's narrative — and recommends keep, remediate, or reject.
Connects to: Frozen acceptance criteria, Evidence custody, Rollback & human authority
evidence-custody
Evidence custody
Hashes, diffs, test output, residue checks, and decisions form a re-auditable record held independently of the agent that made the change.
Connects to: Independent audit, Protected execution, Rollback & human authority
rollback-authority
Rollback & human authority
Every change is reversible, and a human holds the final keep/iterate/reject decision. Authority is never delegated to agent confidence.
Connects to: Scope & path control, Independent audit, Evidence custody
direction
Future direction: Observe, audit, enforce, and orchestrate the change lifecycle without handing authority to the agent that generated the work.