Skip to content
Get packs

Guide · Agent skill

How to use the FluidSim Simulation Planner agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The FluidSim Simulation Planner is an agent skill, a folder of instructions and helper scripts that an AI agent loads when your request matches its purpose. This guide explains what it makes the agent do, how to install it, how to ask for it in natural language, and what it needs to run. It is written for readers who have not used agent skills or FluidSim before.

FluidSim Simulation PlannerK-Dense-AI/scientific-agent-skills · MIT

Quick start

  1. 1Download the ZIP. Get the fluidsim skill ZIP from this page.
  2. 2Unzip into a skills folder. Unzip into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder.
  3. 3Or upload in the Claude apps. Go to Settings → Capabilities → Skills and upload the ZIP.
  4. 4Create a locked Python environment. Use Python 3.11 or newer and install fluidsim[fft]==0.9.0, fluidfft==0.4.5 and pyFFTW==0.15.1 with uv, then lock them.
  5. 5Ask with a physics question. Describe the equations, domain, boundaries and what you want to measure in plain language.
  6. 6Review the plan and validate. Check the resource plan and CFL, resolution and dealiasing bounds, then let the bundled checkers run.
  7. 7Run a tiny pilot. Run one small serial pilot and inspect budgets, spectra and time-step history before scaling up.

What the fluidsim skill makes the agent do

The skill covers FluidSim, a Python framework for numerical solvers, especially periodic Cartesian pseudospectral computational fluid dynamics (CFD). It guides the agent to plan, configure, inspect, restart and analyze bounded simulations.

Before running anything, the agent is steered to state the physical assumptions, select a verified solver, and write a strict resource plan covering CPU, memory, disk, wall time, CFL and resolution limits. It then runs bundled checker scripts. The skill also states plainly that a finished run is not proof of physical validity.

  • State equations, units, geometry, boundaries, initial conditions, forcing and acceptance criteria.
  • Pick a verified solver and inspect its default parameters.
  • Validate the configuration and estimate resources before any run.
  • Generate a dry-run script that does nothing unless executed with an explicit config-ID acknowledgement.
  • Run one tiny serial pilot, then refine grid and time step separately.
  • Prepare, but never automatically submit, a site-specific MPI job.
  • Preserve config, script, uv.lock, versions, logs, output inventory, checksums and restart lineage.

Install the fluidsim skill

Download the skill ZIP from this page and unzip it into one of two locations. For a personal skill available across projects, use ~/.claude/skills/. For a skill that only applies to one project, unzip it into that project's .claude/skills/ folder.

If you use the Claude apps instead, upload the ZIP under Settings → Capabilities → Skills.

  • Personal install: unzip into ~/.claude/skills/
  • Project install: unzip into the project's .claude/skills/
  • Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Set up the FluidSim environment the skill expects

The skill's own scripts use only the Python standard library, but the simulation examples need FluidSim with its FFT extra, plus FluidFFT and pyFFTW. The skill's source gives these pinned versions: fluidsim 0.9.0 with the fft extra, fluidfft 0.4.5 and pyFFTW 0.15.1. Python 3.11 or newer is required.

The source recommends a uv project lock for reproducibility. Direct pins alone do not freeze all dependencies.

  • Python 3.11 or newer
  • fluidsim[fft]==0.9.0, fluidfft==0.4.5, pyFFTW==0.15.1
  • Optional: h5py for HDF5 metadata tools
  • Optional for MPI: a compatible MPI runtime, FFTW/PFFT/P3DFFT libraries, compilers and an approved scheduler
uv init --python 3.12 uv add "fluidsim[fft]==0.9.0" "fluidfft==0.4.5" "pyFFTW==0.15.1" uv lock uv sync --frozen

Trigger the skill with natural requests

You do not need to name the skill or its scripts. Describe the physics question and the simulation task in plain language. The skill's description matches requests about FluidSim solver selection, parameter review, FFT or MPI setup, output diagnostics and restart compatibility.

Be specific about the physical question. The agent will ask for missing assumptions, units, boundary conditions and acceptance criteria rather than guessing them.

  • "Use FluidSim to review my ns2d config for a periodic 2D turbulence run and check the CFL limit."
  • "Estimate CPU, memory and disk for a 256 by 256 ns2d run before I launch anything."
  • "Check whether my restart directory is compatible with the current FluidSim version."
  • "Summarize the output inventory of this FluidSim run folder."
  • "Help me prepare a site-specific MPI job for an ns3d simulation, but do not submit it."

What the bundled scripts do

The skill includes several helper scripts. Each one supports a step in the workflow: validating a plan, estimating grid resources, summarizing a budget, listing outputs and checking restart compatibility. The skill's file list includes budget_summary.py, grid_resource_estimator.py, output_inventory.py and restart_compatibility.py, plus shared helper modules. Reference files cover installation, parameters, solvers, the simulation workflow, output analysis and advanced features.

  • budget_summary.py: summarizes budgets from a run
  • grid_resource_estimator.py: estimates resources for a grid
  • output_inventory.py: inventories output files
  • restart_compatibility.py: checks restart compatibility

Safety notes: scripts and network

The skill runs scripts, and it can read and write files and run shell commands. Review the plan and any generated script before you allow execution. The skill's dry-run step is designed so a script does nothing unless you explicitly acknowledge the config ID.

The skill does not use the network. MPI jobs are prepared for review only; the skill never submits or launches them automatically. The skill does not assume a GPU backend, and GPU use is described as experimental.

  • Runs scripts: yes. Uses network: no.
  • Dry-run scripts require an explicit config-ID acknowledgement to execute.
  • Do not submit MPI jobs without your cluster's approved scheduler workflow.
  • Treat GPU work as experimental and validate it separately.

Tips for useful results

Give the agent your physical question, not only the code. Ask for a tiny serial pilot before a large run. Refine the grid and time step separately so you can see which change affects your observables. Keep the lock file and versions with your results so the run can be reproduced.

  • Start with a small grid and short end time.
  • Check conservation and budget residuals, not only whether the run finished.
  • Keep plots descriptive until convergence checks pass.
  • Record the forced variable, seed and forcing band if you use random forcing.

What people use it for

  • Planning a periodic 2D turbulence run with a reviewed CFL limit and output size estimate.
  • Checking a FluidSim run folder's output inventory and summarizing its budgets.
  • Confirming whether saved restart files are compatible before resuming a simulation.
  • Preparing a reviewed, site-specific MPI job description without submitting it.

Frequently asked questions

How do I install the fluidsim agent skill?

Download the skill ZIP from this page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. In the Claude apps, upload the ZIP under Settings → Capabilities → Skills.

What do I need to run the fluidsim skill?

The skill's scripts use the Python standard library and require Python 3.11 or newer. Simulation examples target fluidsim 0.9.0 with the fft extra, fluidfft 0.4.5 and pyFFTW 0.15.1. MPI use needs a compatible MPI runtime, FFT libraries and an approved scheduler.

Is the fluidsim skill free, and what is its licence?

The skill is released under the MIT licence, as stated in its metadata. The skill's source notes that upstream FluidSim is licensed under CeCILL-2.1, and that the MIT licence applies only to the skill.

Does the fluidsim skill need an internet connection?

No. The skill's metadata states it does not use the network. It does run scripts, so review the plan and any generated script before allowing execution.

Does the skill support GPU simulations?

No GPU backend is assumed. The source says GPU use is experimental and requires separate validation, so the skill does not claim GPU acceleration.

Does a finished simulation mean my results are correct?

No. The skill states that a completed run, a stable time step, a smooth plot or a closed program exit does not prove convergence or physical validity. You still need to check resolution, conservation and grid and time convergence.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

This skill suits researchers and engineers who plan periodic Cartesian pseudospectral fluid simulations in Python and want reproducible configs and reviewed resource estimates. It is a poor fit for anyone wanting a quick GPU-accelerated run, a guaranteed-correct answer, or who lacks a physics question and a Python environment.

Download the skill