confidence-check-skills

Pre-implementation validation uses a weighted 5-factor model to evaluate duplicates, architecture alignment, documentation, OSS references, and root-cause analysis before coding. This framework prevents wasted development effort by ensuring high-confidence plans for features, refactors, and architecture changes.

To check for code duplication and architecture alignment before adding a feature, run a 5-factor confidence assessment using Glob, Grep, Read, and WebSearch. This validates your plan against existing code and documentation to prevent wasted development effort.

To decide whether to proceed, clarify, or stop a software refactor, apply the 0.90 and 0.70 confidence decision thresholds to your 5-factor assessment results. A score above 0.90 means PROCEED, while lower scores trigger CLARIFY or STOP to prevent wasted effort.

Yes, this pre-implementation validation framework integrates with optional semantic search using ChromaDB and TodoWrite. It also leverages Glob, Grep, Read, and WebSearch to execute the weighted 5-factor assessment for architecture changes and refactors.

A limitation of using a quantified confidence check for architecture changes is that it requires executing a weighted 5-factor assessment before coding. If confidence scores fall below the 0.70 threshold, you must STOP or CLARIFY, which may delay immediate implementation.