llm-challenge
Installation
SKILL.md
LLM Challenge
Benchmark for @tailor-platform/sdk AI-friendliness. Located at llm-challenge/.
Read llm-challenge/README.md for commands, scoring, and verification details.
Core Rule
When AI fails a challenge, improve the SDK (JSDoc, error messages, types, CLAUDE.md) — NEVER add hints to problem.md.
Prerequisite
ALWAYS build SDK before running: pnpm -C packages/sdk build
Problem Conventions
Structure: problems/<id>-<name>/ with meta.json, problem.md, scaffold/, solution/, tests/
meta.json rules:
Related skills