# P04 Preregistration

## Title

Local Green, Global Red: DAG-of-DAG Detection of False Stability and Correction Residue

## Claims

- `P04-C1`
- `P04-C2`
- `P04-C3`
- `P04-C4`

## Dependencies

Frozen parent foundation only.

## Primary outcomes

- False completion rate
- Broken-join detection
- Correction completeness
- Stale residue
- Convergence time
- Coordination overhead
- False-edge cost
- Availability

## Controls

- Flat dashboard
- Manual tracing
- Ordinary dependency DAG
- Rebuild-all
- No suspected-edge discovery
- No human-memory graph
- No lifecycle separation
- Perfect dependency oracle upper bound

## Analysis plan

Results are stratified by dependency completeness, graph size, change depth, and human-memory involvement. Pareto analysis reports completeness versus overhead. Hidden-edge inference is evaluated as a precision-recall problem separate from correction. We preregister an adverse region in which rebuild-all should dominate because the graph is too incomplete or maintenance cost too high. No aggregate score is permitted to hide a method that is cheap but leaves critical residue.

## Null handling

If ordinary dependency tracing matches correction completeness at lower cost, the meta-graph claim is not promoted. Suspected edges that cannot be confirmed remain suspected and are not counted as true discoveries. A no-residue scan after a bounded observation window is not treated as proof of completeness. Missing ground truth is reported as unknown rather than negative.

## Falsifiers

- No reduction in false readiness
- Correction completeness matches simple DAG
- Suspected-edge false positives exceed benefit
- Graph maintenance erases recovery gains
- Lifecycle separation adds no predictive value

## Release boundary

This public paper is limited to benign assurance, provenance, human-agency preservation, lawful state transition, simulation, test infrastructure, and defensive resilience. It does not disclose operational target-selection logic, engagement optimization, platform-specific exploitation thresholds, live deception procedures, signature-emitter recipes, or methods that materially increase harmful capability. Any later operational research requires separate lawful authority, ethics and safety review, configuration control, and release adjudication.

## Deviations

Any deviation from the frozen design must be timestamped, justified, and separated into an exploratory analysis. No favorable result may retroactively alter the primary gate.