Enforce JIT-safe numeric-only Numba functions in src/argus/analysis.
npx skills add https://github.com/lagarcess/argus --skill numba-patterns
Please help me install this Agent Skill. Skill: Numba Patterns Source: https://github.com/lagarcess/argus/tree/main/.agent/skills/numba-patterns Command: npx skills add https://github.com/lagarcess/argus --skill numba-patterns
This Skill codifies rules and best practices to ensure Numba JIT-compiled functions in src/argus/analysis are pure numeric, deterministic, and meet strict performance budgets so developers avoid hidden compilation costs and non-JIT-safe patterns that break benchmarks.
Run the Numba Patterns guidance to check that all @njit functions in src/argus/analysis are pure-math, that tests call warmup_jit() before timing, and that wrapper functions convert raw arrays into Python objects.
High-intent search queries and answers about installing and using this skill.
Numba JIT functions require strict numeric-only operations, prohibiting Python objects and I/O inside @njit or @jit(nopython=True) decorated routines to guarantee deterministic, high-performance compilation without runtime type errors.
Numba benchmark tests must call warmup_jit() in the test setup to separate JIT compilation overhead from actual runtime metrics, ensuring you measure true execution speed for large datasets.
Use a wrapper pattern where lightweight Python wrapper functions convert raw JIT arrays output from @njit functions into higher-level objects, keeping the JIT routines pure numeric while enabling object-oriented analysis.
Achieve sub-50ms post-warmup Numba performance by enforcing JIT-safe numeric patterns, avoiding forbidden Python libraries inside @njit functions, and validating pivot detection benchmarks after calling warmup_jit() to exclude compilation overhead.
Forbidden operations inside Numba @njit functions include using Python objects, performing I/O operations, and importing unsupported libraries, which breaks JIT compilation and prevents deterministic high-performance numeric computation.
Python objects are prohibited inside Numba JIT functions to maintain JIT safety; pivot detection routines must use pure numeric math within @njit decorators to ensure deterministic execution and meet strict performance budgets.
KonjkovCompile numerical Python functions with Numba JIT decorators.
lagarcessEnforce pytest best practices and TDD workflows for Argus analysis engine tests.
x-cmdMeasure command execution time with microsecond precision and analyze statistics.
jkitchinProvide expert guidance for JAX numerical computing with grad, jit, vmap, and pmap.
