Tests idiomatic Python code against CPython and Monty to detect behavioral divergences.
npx skills add https://github.com/pydantic/monty --skill review-usability
Please help me install this Agent Skill. Skill: review-usability Source: https://github.com/pydantic/monty/tree/main/.agents/skills/review-usability Command: npx skills add https://github.com/pydantic/monty --skill review-usability
When developing the Monty Python interpreter, changes on a branch can silently break common Python idioms that LLMs typically write, causing wrong results instead of clean errors. This Skill systematically compares branch behavior against CPython to surface divergences before they reach users.
Review the current branch for usability by testing common Python idioms in the playground against both CPython and Monty, then report any divergences you find.
High-intent search queries and answers about installing and using this skill.
Write real test files exercising the idioms a model would use first, then run each under both CPython (uv run) and your interpreter (cargo run) and diff the outputs. Focus on the obvious methods and argument forms rather than exotic edge cases.
Differential testing runs identical code under two implementations, here CPython and Monty, and compares results. Divergences reveal bugs, with silent wrong answers prioritized over clean errors because users cannot recover from incorrect output.
A missing feature that raises AttributeError is recoverable because the caller sees the failure and can adapt. Silent divergence returns a wrong answer with no signal, so the model has no way to know it must write different code.
No, the review is report-only unless the user explicitly asks for fixes. It leaves playground test files in place and documents each divergence with the code, both outputs, and likelihood of occurrence.
Test the most common patterns an LLM would plausibly write: obvious method calls, standard argument forms, and typical builtin combinations. Exotic corners matter less because real usage concentrates on ordinary idiomatic code.
mozillaDetect code smells in files changed on the current branch against main.
elasticFlag internal jargon and abbreviations in documentation files with plain-language replacements.
rust-worksIdentify and address common testing anti-patterns across software projects.
sentenzAutomate fuzz testing of Go packages with native fuzzing engine.
