udales-detect

name	udales-detect
description	Use when users ask how to install, build, run, or test uDALES on this machine (Windows/WSL/Linux/HPC), what modules/toolchain to use, or how to detect the current environment before execution; gather environment fingerprints and route execution tasks to udales-exec.

uDALES Detection Skill

This skill identifies the execution environment, gathers system fingerprints, checks for site documentation, and determines which uDALES workflow is needed.

When to use

Use this skill when a user asks:

• "How do I run this here?"
• "What modules do I need?"
• "Which MPI/Python stack should I use on this system?"
• "How do I install preprocessing on my machine?"
• "Should I use Windows, WSL, local Linux, or cluster modules for this?"
• "Can you detect this environment first and then give run/build steps?"

Steps

1. Run the local probe to capture a baseline fingerprint.
2. Check known cluster profiles in references/clusters.md.
3. If no match, query online documentation for the host/cluster.
4. Ask what the user needs:
   • preprocessing and postprocessing
   • solver execution
   • both
5. Based on the answer, direct execution work to the udales-exec skill.
6. Record new details in references/clusters.md.

Local probe

Run:

.github/skills/udales-detect/scripts/detect_env.sh

This prints hostname, OS, module availability, mpiexec flavor, and Python paths.

Online lookup

If the cluster is unknown, search for:

• cluster name or hostname suffix (e.g. *.imperial.ac.uk)
• module environment documentation
• MPI launcher guidance
• Python module guidance

Record any relevant links in references/sources.md.

Minimal checks

Only run lightweight checks during detection:

• mpiexec --version
• python3 --version

Leave build/run checks to udales-exec after the user confirms the intent.

Files to consult

• docs/cluster_workflows.md
• tools/build_executable.sh
• tools/build_preprocessing.sh
• tests/README.md
• .github/skills/udales-exec/references/clusters.md