Generate, evaluate, and refine SKILL.md files using teacher-student model distillation. Built in Rust.
cargo install thulpoff-cliA capable teacher model solves a task. Thulpoff extracts the solution patterns into a reusable SKILL.md file with test cases.
Run test cases against student models with and without the skill context. Baseline comparison shows exactly how much the skill helps.
Failed tests feed back to the teacher model, which improves the skill definition. The loop continues until the student passes.
| Crate | What |
|---|---|
thulpoff-core | Types, traits, LlmProvider |
thulpoff-provider | AnthropicProvider, NimProvider |
thulpoff-engine | Generation, Evaluation, Refinement + baseline + history |
thulpoff-cli | generate, eval, refine, list, runs |
# Generate a skill from a teacher session
thulpoff generate "Write an optimized sorting algorithm" --model claude-opus-4-6 --provider anthropic
# Evaluate with a student model
thulpoff eval ./skills/sorting/SKILL.md --model mistralai/mistral-small-24b-instruct-2501 --provider nim
# View run history
thulpoff runs sorting
# Refine based on failures
thulpoff refine ./skills/sorting/SKILL.md --model claude-opus-4-6 --provider anthropic