claude-desc-skills

Ecosystem-level guide to the DESC (Dark Energy Science Collaboration) Python stack. Covers which tools exist, how they fit together, general DESC coding conventions, and how to set up the shared DESC Python environment and Jupyter kernels at NERSC. Use this skill when the user is new to DESC or asks what DESC software exists; asks how DESC packages fit together; asks about cross-package data flow (e.g. TXPipe → sacc → Firecrown, or RAIL inside TXPipe); asks about general DESC coding conventions; needs to choose the right tool for a DESC analysis task; asks how to activate or install the DESC Python environment; asks about desc-python, desc-python-dev, desc-python-bleed, or other DESC Jupyter kernels; or asks a question spanning multiple DESC packages with no single obvious skill to invoke. For questions specific to a single package (ceci, TXPipe, RAIL, pyccl/CCL, pymaster/NaMaster, firecrown, photerr), prefer that package's dedicated skill instead.

Guide to photerr, a Python library for photometric error modeling in astronomical imaging surveys (LSST/Rubin, Euclid, Roman). Covers ErrorModel/ErrorParams architecture, survey-specific models (LsstErrorModel, EuclidErrorModel, RomanErrorModel and variants), parameter customization (nYrObs, nVisYr, m5, theta, gamma, scale, sigmaSys), non-detection handling (ndMode, sigLim, ndFlag), extended-source aperture corrections (extendedType="auto"/"gaap"), and getLimitingMags. Use this skill whenever the user imports photerr; works with LsstErrorModel, EuclidErrorModel, RomanErrorModel, or ErrorModel; calls getLimitingMags; simulates photometric errors or survey depth; applies photometric noise to a magnitude catalog; handles non-detections or limiting magnitudes; models extended-source (galaxy) photometry; compares survey depths or error budgets across LSST, Euclid, or Roman; or asks how to add realistic photometric scatter to a simulated catalog. For RAIL pipeline stages that wrap photerr (LSSTErrorModel degradatio

How to write DESC pipeline stages and pipelines with ceci. Covers PipelineStage subclassing, inputs/outputs declarations, config_options, HDFFile/FitsFile/TextFile/YamlFile I/O, open_input/open_output, parallel iteration helpers, MPI patterns, pipeline YAML structure (modules/launcher/site/stages/inputs), stage aliasing, CLI usage, resume mode, nprocess, and NERSC/parsl deployment. Use this skill whenever the user imports ceci; subclasses PipelineStage; defines inputs, outputs, or config_options on a stage; calls open_input or open_output; writes or debugs a pipeline YAML; runs ceci from the CLI; asks about stage aliasing, DuplicateStageName, resume mode, nprocess/MPI, parsl launchers, or ceci file types.

Guide to NaMaster (pymaster), the DESC pseudo-Cl angular power spectrum estimator. Use this skill whenever the user imports pymaster; works with NmtField, NmtBin, NmtWorkspace, or NmtCovarianceWorkspace; calls compute_full_master, compute_coupled_cell, decouple_cell, or mask_apodization; computes pseudo-Cl power spectra for spin-0 or spin-2 (shear) fields; handles masks, mode-coupling matrices, or workspace serialization; asks about bandpower binning, EE/BB/EB spectrum ordering, or apodization; writes TXPipe Fourier-space stages (TXTwoPointFourier, TXFourierNamasterCovariance); or needs Gaussian covariance estimates.

How to work with RAIL (Redshift Assessment Infrastructure Layers), the DESC photo-z framework built on ceci. Covers RailStage/DataStore/DataHandle patterns, namespace architecture (rail.* entry-points), DataHandle types (QPHandle, PqHandle, Hdf5Handle, ModelHandle), three modules (rail.creation/estimation/evaluation), CatInformer/CatEstimator/PZEstimator base classes, RailPipeline/RailProject, pipeline YAML format, and the catalog simulation workflow (QuantityCut/Reddener/LSSTErrorModel/Dereddener/SpecSelection). Use this skill whenever the user imports rail; works with RailStage, DataStore, or any DataHandle type; does photo-z training, estimation, or evaluation; uses any rail.creation degradation stage (QuantityCut, Reddener, LSSTErrorModel, Dereddener, ObsCondition, SpecSelection_DESI_Phy); works with qp.Ensemble PDFs; sets up a RailPipeline or RailProject; asks about photo-z algorithms (FlexZBoost, BPZ, KNN, GPz, DNF, etc.); or asks about the rail.* namespace and entry-point architecture.

Best-practices guide for designing and auditing Claude Code skills and global configuration. Covers CLAUDE.md length and content fitness, skill naming and description quality, SKILL.md structure and progressive disclosure, trigger condition robustness, settings.json permissions, and directory hygiene. Use whenever asked to write, update, audit, or improve a SKILL.md file; evaluate trigger condition quality; review or suggest improvements to a CLAUDE.md or settings.json; create a new skill from scratch; improve an existing skill; or assess whether skills in a set cross-reference each other correctly.

How to work with TXPipe, the DESC 3x2pt weak-lensing pipeline. Covers all pipeline phases (ingestion → photo-z → selection → calibration → masks → maps → noise maps → 2pt → covariance → theory → diagnostics), key stage names, file tags, pipeline YAML structure, and critical gotchas. Use this skill whenever the user imports txpipe; works with any TXPipe stage (TXTwoPoint, TXTwoPointFourier, TXSourceMaps, TXLensMaps, TXSourceSelector*, TXShearCalibration, TXTracerMetadata, TXSimpleMask, TXCustomMask, TXIngest*, TXBlinding, TXFourierNamasterCovariance, TXRealNamasterCovariance, TXJackknifeCenters, TXPSFDiagnostics, TXRoweStatistics, TXTauStatistics, etc.); writes a TXPipe pipeline YAML; asks about shape or shear catalogs, metacalibration or metadetect, source or lens selection, tomographic binning, 3x2pt two-point functions (shear-shear, galaxy-galaxy lensing, angular clustering), map generation, noise maps, jackknife covariance, PSF diagnostics, or theory predictions.

Guide to CCL (pyccl), the DESC Core Cosmology Library for theoretical cosmological calculations. Use this skill whenever the user imports pyccl, works with Cosmology objects, computes angular power spectra (angular_cl), uses tracers (WeakLensingTracer, NumberCountsTracer, CMBLensingTracer), computes 3D power spectra (linear_matter_power, nonlin_matter_power), uses Pk2D, calls scalar functions like sigma8/growth_factor/comoving_radial_distance, configures halo model components, or asks about CCL cosmological calculations, units conventions, or parameter naming in any DESC pipeline context.

Guide to Firecrown, the DESC likelihood framework for cosmological parameter inference. Covers the Likelihood/Statistic/Systematic/ModelingTools architecture, two-point statistics (angular power spectra and correlation functions), systematic effects (intrinsic alignment, photo-z shifts, multiplicative shear bias, galaxy bias), sacc as the data container interface, and connectors to CosmoSIS, NumCosmo, and Cobaya. Also covers the TXPipe→Firecrown data flow via sacc files. Use this skill whenever the user imports firecrown, works with likelihood objects, builds a Firecrown analysis, connects to CosmoSIS or NumCosmo, reads or writes sacc files for inference, asks about DESC likelihood inference, or calls `load_likelihood` / `build_likelihood`.

How to work productively at NERSC. Covers Slurm job submission and QOS selection (salloc/sbatch/srun, --account, --qos, --constraint), filesystem strategy (HOME/CFS/PSCRATCH/COMMUNITY) and quotas, login-vs-compute discipline, NERSC-specific gotchas (mandatory --account, GPU repo `_g` suffix, $PSCRATCH purge policy, srun for everything that runs on compute, login-node CPU caps), and module/conda environment setup. The current flagship system is Perlmutter (CPU = AMD Milan, GPU = 4× A100/node); the durable conventions carry across system upgrades. DESC NERSC repos: m1727 (CPU), m1727_g (GPU). Use this skill whenever $NERSC_HOST is set, the user runs Slurm commands, touches /global/cfs or /pscratch, works with module load, or asks about NERSC-specific workflows.