Maths
Install this skill (skills CLI)
Add this skill to your agent from the skills ecosystem:
# List skills in this repo
npx skills add udnisap/skills --list
# Install the maths skill (project scope)
npx skills add udnisap/skills --skill maths -y
# Install globally for all projects
npx skills add udnisap/skills --skill maths -g -yRepo: github.com/udnisap/skills.
Setup (do this first)
Use a virtual environment and install the maths libraries inside it. Step-by-step: install.md. Install by OS (Linux, macOS, Windows): install-by-os.md.
- Check (with venv activated):
python -c "import numpy, scipy, sympy, mpmath; print('OK')"— if this fails, follow install.md or install-by-os.md for your platform. - Quick setup: create venv (e.g.
python3 -m venv.venv), activate it, thenpip install numpy scipy sympy mpmath. Run scripts with that venv’spython.
Use established Python libraries instead of hand-rolled math. Choose by task:
| Need | Library |
|---|---|
| Arrays, linear algebra, FFT, random | NumPy |
| Optimization, integration, ODEs, stats, sparse | SciPy |
| Symbolic math, simplify, solve, differentiate | SymPy |
| Arbitrary-precision floats, special functions | mpmath |
When to use
- Numerical arrays, matrix ops, eigenvalues, SVD, FFT
- Minimization, root finding, curve fitting, integration, ODEs
- Symbolic expressions, equation solving, calculus (symbolic)
- High-precision decimals or special functions beyond float64
Quick workflow
- Identify numeric vs symbolic vs high-precision.
- Import only the submodule you need (e.g.
scipy.optimize,sympy.solvers). - Prefer library functions over custom loops (vectorize with NumPy; use
scipy.integrate,sympy.integrate, etc.).
Using from the CLI (for agents)
When the agent must run maths from the command line, use Python inline:
- One-liner:
python -c 'import numpy as np; print(np.linalg.det([[1,2],[3,4]]))' - Multi-line: use a single string with semicolons, or write a short script and run
python script.py. For longer code, prefer a temp file over a huge-cstring. - SymPy (expression in, result out):
python -c "import sympy as sp; x=sp.Symbol('x'); print(sp.solve(x**2-4,x))" - Exit code: script should
print()the result and exit 0; the agent reads stdout. Usesys.exit(1)on error so the agent can detect failure.
Use the venv’s python (or python3); ensure the venv has the needed packages (see install.md).
Scripts
Runnable examples are in scripts/. With the venv activated, run them from the skill directory:
scripts/numpy_solve.py— solve Ax = b (option:--eigfor eigenvalues)scripts/scipy_minimize.py— minimize a scalar functionscripts/scipy_integrate.py— definite integral (Gaussian)scripts/sympy_solve.py— solve equation (arg: expression, e.g."x**2 - 4")scripts/sympy_integrate.py— symbolic integral (arg: expression, e.g."sin(x)**2")scripts/mpmath_precision.py— high-precision integral (optional arg: dps)
See examples.md for exact CLI commands and one-liners.
Additional resources
- For environment setup (venv and library install), see install.md; for OS-specific install, see install-by-os.md.
- For API patterns and code snippets per library, see reference.md.
- For worked examples and CLI usage, see examples.md.