Astropy Repo Skill

Use this skill when a task needs the Astropy core Python package for astronomy or astrophysics analysis. Astropy is a CPU scientific-Python library; selected workflows do not require GPU or accelerator backends. Use an installed Astropy package and the bundled references/scripts here rather than reopening the source repository.

Fast Start

import astropy
from astropy import units as u
print(astropy.__version__)
print((42 * u.km / u.s).to(u.m / u.s))

If plotting, fitting, convolution FFT paths, or common dataframe bridges are needed, prefer installing the common optional set:

python -m pip install "astropy[recommended]"

Use astropy[all] only for broad optional integrations such as HDF5, Parquet, S3/fsspec, Jupyter, or additional astronomy packages.

Route by Task

• For units, quantities, equivalencies, custom units, and physical constants, read sub-skills/units-constants/SKILL.md.
• For Time, timescales, SkyCoord, coordinate frames, transformations, separations, matching, and observation geometry, read sub-skills/time-coordinates/SKILL.md.
• For Table/QTable, unified I/O, FITS, ASCII/ECSV, VOTable, HDF5/Parquet notes, pandas/dataframe bridges, and table/FITS CLIs, read sub-skills/tables-io/SKILL.md.
• For FITS WCS, pixel/world conversion, WCS validation, NDData, CCDData, masks, uncertainties, and image metadata containers, read sub-skills/wcs-nddata/SKILL.md.
• For image normalization, intervals, stretches, WCSAxes, RGB output, fits2bitmap, kernels, direct convolution, and FFT convolution, read sub-skills/visualization-convolution/SKILL.md.
• For models, fitters, compound models, constraints, robust statistics, histograms, time series, periodograms, and uncertainty distributions, read sub-skills/modeling-stats-timeseries/SKILL.md.
• For cosmology realizations/classes, distances, ages, redshift inversion, cosmology units, and serialization, read sub-skills/cosmology/SKILL.md.
• For install/import checks, optional extras, config/cache, remote data/IERS, logging/warnings, SAMP, and the general CLI catalog, read sub-skills/cli-config-data/SKILL.md.

Shared References and Scripts

• Read references/package-overview.md for the module map, install extras, and general workflow pattern.
• Read references/troubleshooting.md for cross-cutting install/import, optional dependency, remote-data, units, and FITS/WCS safety guidance.
• Read references/repo-provenance.md to check the source commit and evidence baseline before refreshing this skill.
• Router import metadata lives in references/repo-routing-metadata.json.
• Run scripts/astropy_smoke.py to verify a small installed-package API surface across the main sub-skill routes.
• Run scripts/astropy_cli_smoke.py to verify public Astropy console commands with safe help and optional temporary fixtures.

Operating Rules

1. Keep units attached until a non-Astropy API requires raw arrays; then use .to_value(unit) and record the unit.
2. Be explicit about format=, scale=, frame=, unit=, origin=, and format names for I/O to avoid silent guessing.
3. Use temporary files for smoke checks and never mutate user FITS/table files unless the user explicitly requested that exact operation.
4. For offline or reproducible tasks, disable or control remote downloads before coordinate/time/IERS/name-resolution operations.
5. Validate scientific transformations numerically: unit round-trips, pixel/world round-trips, table read/write round-trips, fit residuals, or cosmology unit checks.
6. Route optional dependency errors to the nearest sub-skill first, then to the root troubleshooting reference for install-level decisions.

Do Not Use This Skill For

• General NumPy/Pandas data processing without Astropy APIs, astronomy formats, units, times, or coordinates.
• Maintaining the Astropy repository, release engineering, CI, or documentation build workflows.
• Running broad Astropy test suites or benchmarks as a substitute for a focused user workflow.
• SAMP services, remote data downloads, or file-mutating CLI operations without explicit user approval and bounded temporary-file safeguards.