numerical-patterns

Numerical computing patterns for C++20 including matrix operations, iterative solvers, numerical stability, data pipelines, and HPC I/O with MPI-IO and HDF5.

Install with Tessl CLI

npx tessl i github:ysyecust/everything-claude-code --skill numerical-patterns

What are skills?

Quality

64%

Does it follow best practices?

Impact

Pending

No eval scenarios have been run

Optimize this skill with Tessl

npx tessl skill review --optimize ./skills/numerical-patterns/SKILL.md

SKILL.md

Review

Evals

Evaluation results

90%

Floating-Point Accumulation Utilities

Numerical stability algorithms

Criteria

Without context

With context

Kahan compensation variable

100%

std::span parameter for sum

100%

StableL2Norm max scaling

100%

Zero max_val guard in norm

100%

std::span parameter for norm

100%

Catastrophic cancellation test

100%

Machine epsilon edge case

Overflow/underflow norm test

100%

PASS/FAIL output and exit code

100%

C++20 compilation

100%

Empty input guard

100%

Without context: $0.3007 · 8m 36s · 15 turns · 162 in / 4,065 out tokens

With context: $0.4855 · 13m 41s · 20 turns · 396 in / 5,600 out tokens

100%

16%

Sparse Linear System Solver for Heat Diffusion

Iterative solvers and matrix patterns

Criteria

Without context

With context

Row-major DenseMatrix

100%

std::span data accessor

100%

std::span in Apply

42%

100%

CSR sparse format

100%

SolverResult struct

100%

Default tol = 1e-10

100%

Default max_iter = 10000

100%

Relative residual convergence

100%

Breakdown detection

50%

100%

Templated matrix type

100%

C++20 compilation flag

100%

Assertion on dimensions

100%

Without context: $0.5171 · 14m 7s · 18 turns · 145 in / 8,619 out tokens

With context: $1.1586 · 19m 14s · 32 turns · 298 in / 16,794 out tokens

91%

52%

Scientific Simulation Data Processing Pipeline

Data pipeline and binary I/O

Criteria

Without context

With context

DataPipeline AddStage chaining

100%

Stage function signature with span

100%

std::ranges::max_element usage

std::ranges::transform usage

100%

std::ranges::copy_if usage

100%

Binary file I/O mode

100%

Default chunk size 1<<20

100%

reinterpret_cast for raw double I/O

100%

Pipeline Process method

50%

100%

C++20 compilation flag

100%

Template on element type

100%

Without context: $0.8769 · 15m 8s · 27 turns · 217 in / 14,019 out tokens

With context: $0.8796 · 14m 50s · 22 turns · 169 in / 14,601 out tokens

Evaluated: 5 days ago
Agent: Claude Code
Model: Claude Sonnet 4.6

Table of Contents

Floating-Point Accumulation Utilities Sparse Linear System Solver for Heat Diffusion Scientific Simulation Data Processing Pipeline

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.