Compute 100-point quality scores for NL programming artifacts using predetermined penalties.
npx skills add https://github.com/xiaolai/nlpm-for-claude --skill scoring-xiaolai
Please help me install this Agent Skill. Skill: scoring Source: https://github.com/xiaolai/nlpm-for-claude/tree/main/skills/nlpm/scoring Command: npx skills add https://github.com/xiaolai/nlpm-for-claude --skill scoring-xiaolai
The scoring system provides a consistent, objective way to evaluate NL programming artifacts by applying penalties and calibrating judgments using a 100-point rubric.
Use cases: quality audits of NLPM components across repos, onboarding evaluators, CI checks.
Run the NLPM score command to compute 100-point artifact scores and penalties for all NL artifacts.
High-intent search queries and answers about installing and using this skill.
To score NL programming artifacts consistently, apply a deterministic 100-point rubric with type-specific penalties and use calibration examples to anchor judgments. This enforces objective evaluation across all NLPM components.
A deterministic 100-point rubric for NL artifact quality audits is a scoring system that applies predefined penalties for specific issues across artifact types, clamping results between 0 and 100. Calibration examples anchor judgments to maintain consistency.
Yes, you can use automated scoring for CI checks on prompts and memory files. The scoring system computes consistent quality scores by applying predetermined penalties across all NLPM artifact types, including skills, agents, commands, rules, and hooks.
To apply type-specific penalties when evaluating skills and agents, run the NLPM score command. It automatically calculates penalties based on the artifact type and clamps the final quality score to a strict 0-100 range.
Yes, the NLPM scoring system supports calibration examples for onboarding evaluators. These worked examples anchor judgments and improve consistency across reviewers by demonstrating how predetermined penalties apply to various artifact types.
The limitation of using a penalty-based scoring system for NL artifacts is that it enforces a strict 100-point scale with predetermined penalties, meaning scores are clamped at 0-100 and may not capture nuances outside the predefined rubric criteria.
xiaolaiAutomate test-driven development workflows for NL artifacts using .nlpm-test spec files.
Kling0012Automate verification checks with truth scoring and rollback for CI/CD pipelines.
xiaolaiEnforce R01–R50 quality rules on NL programming artifacts.
wangyouanEvaluate agent system performance using deterministic checks and multi-dimensional rubrics.
