What problem does it solve?
ACE-built Deliver apps often perfectly match a thin PDD's structural skeleton but remain undeployable in production due to missing data validation, unenforced GPS accuracy, absent case property writes on follow-up forms, or missing required language translations. This skill eliminates that gap by grading apps not just on PDD conformance, but on real-world deployability against a CommCare specialist's ship-ready bar, catching silent defects that would cause costly field failures post-handoff.
Core Features & Use Cases
- Dual-Axis Grading: Evaluates apps on 9 weighted dimensions across two axes: 5 conformance dimensions (45% total weight) checking alignment with the PDD's field count, question order, consent gate semantics, conditional logic, and Connectify wiring, plus 4 fitness dimensions (55% total weight) assessing capture reliability, data quality enforcement, case persistence, and localization compliance.
- HITL Stub Detection: Automatically detects incomplete human-in-the-loop pending app builds and returns an incomplete verdict instead of generating misleading scores for non-existent app structures.
- Calibrated Actionable Reporting: Produces a structured YAML verdict and human-readable report with specific Nova edit suggestions to resolve identified gaps, calibrated against known ground-truth build pairs to ensure consistent, reliable scoring.
- Use Case: For ACE operators building CommCare Deliver apps for Connect public health opportunities, this skill ensures the final app is not just spec-compliant but ready for real-world field deployment, preventing costly rework after handoff to implementation teams.
Quick Start
Invoke the pdd-to-deliver-app-eval skill to evaluate the Nova-built Deliver app for the active Connect opportunity against its PDD specification, producing a structured conformance and fitness verdict with actionable improvement recommendations.
