python-skill

Configure and run mutmut v3 for Python mutation testing — fork-based execution, `[tool.mutmut]` in `pyproject.toml`, the `mutmut browse` TUI, `# pragma: no mutate` suppression, and the discipline of promoting surviving mutants to new tests. Use after `python-verification` confirms mutmut is the right adversary.

Run symbolic execution on Python functions with CrossHair to find counter-examples to contracts or assertions, fill coverage holes via `cover`, and detect behavioural drift via `diffbehavior`. Use after `python-verification` confirms CrossHair is the right adversary.

Write and maintain Hypothesis property-based tests for Python, including strategy design, the filtering trap, stateful testing with RuleBasedStateMachine, the example database, and CI tiering. Use after `python-verification` confirms Hypothesis is the right adversary.

Use for Python decorators, descriptors, context managers (function and class form), metaclasses, and multiple dispatch. The common question is "which abstraction hides this complexity at the right level?".

Use for choosing between threads, `asyncio`, `multiprocessing`, and PEP 734 subinterpreters in Python. Covers workload shape (CPU vs I/O), shared state and the GIL, cancellation, and the `concurrent.interpreters` / `InterpreterPoolExecutor` API new in Python 3.14.

Use for choosing a Python data container — `msgspec.Struct`, `dataclasses.dataclass`, `attrs`, `TypedDict`, `NamedTuple`, frozen vs mutable, tagged unions, and the boundary between wire format and domain object.

Use for Python exception design, raising, catching, and logging. Covers narrow `except`, `raise … from …`, domain hierarchies with `Error` suffixes, parameterised logging, `logger.exception`, and the Ruff TRY/BLE/EM/LOG/N818/PERF203 rules.

Use for Python iterators, generators, lazy evaluation, and refactoring deeply nested loops by extracting an iterator or a context manager. Covers `yield`, `yield from`, generator expressions, `itertools` discipline, and async iteration.

Use for Python code-quality scanners and profilers beyond unit testing and type-checking — `deadcode` for unused symbols, `pyscn` for unreachable code, clone, and coupling metrics, and Pyinstrument for low-overhead profiling.

Route Python work to the smallest useful skill. Use for Python coding, design, typing, errors and logging, decorators and other abstractions, iterators and generators, data shapes, concurrency, testing, verification, and quality-tool selection.

Use for advanced pytest usage — fixture scopes, parametrization, marks, plugins, snapshot and approval tests, async tests, and the boundary between fast unit, behavioural, and verification testing.

Use for Python typing decisions — generics, `TypeVar`, `ParamSpec`, `TypeIs` and `TypeGuard`, `NewType`, `Protocol`, `@overload`, typed kwargs, and the shape of public function signatures.

Select between Hypothesis, CrossHair, and mutmut for Python verification work. Use when example-based tests are not enough and the question is which adversary should generate or evaluate cases.