Skip to main content

score-quality

Score code migration quality based on coverage and correctness. Evaluates how well target code represents source functionality.

Informações da origem

Repositório
OpenHands/extensions
Última atividade na origem
6 de março de 2026 às 18:36
Idioma detectado do SKILL.md
inglês
Estrelas
157
Forks
102

Opções de instalação

Por padrão, está selecionado o prompt que primeiro revisa a origem. Você pode mudar para um comando direto ou baixar uma cópia local.

Revise os arquivos de origem

Leia o SKILL.md e os arquivos complementares exibidos pelo SkillsMP antes de decidir se vai instalar.

Explorador de arquivos
2 arquivos

Exibindo SKILL.md

SKILL.md
Instruções da origem · Visualização somente leitura
name
score-quality
description
Score code migration quality based on coverage and correctness. Evaluates how well target code represents source functionality.
license
MIT
compatibility
Requires completed migration with source and target code
triggers
["migration quality","migration score","coverage score","correctness score"]
Evaluate migration quality by scoring coverage and correctness for each source-target file pair. **Prerequisite**: A migration mapping file must exist (see `migration-mapping` skill). See [references/scoring-criteria.md](references/scoring-criteria.md) for the 1-5 scoring scales. ## Output Format Save scores as a JSON file: ```json { "source_file_1.cbl": { "coverage": 4, "correctness": 5, "justification": "All calculation logic migrated. Tax rounding matches COBOL behavior." } } ``` ## Evaluation Process For each source file and its mapped target files: 1. **Identify key functionality** in the source - Business logic, data operations, control flow, error handling 2. **Check coverage** in the target - Is each function/paragraph represented? - Are all data fields and code paths covered? 3. **Verify correctness** - Do calculations produce the same results? - Are edge cases handled the same way? 4. **Document justification** - Specific functionality present/missing - Known discrepancies ## Incremental Scoring If a score file already exists, update it with new scores rather than replacing it entirely.
Ver no GitHub