Fluent CFD Skill
A Codex skill for Ansys Fluent and PyFluent workflows: setup, boundary conditions, convergence and safe MCP use
- 90 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The skill directs the agent to run Fluent and TUI commands through MCP, and calculations spend a license and compute time
- Arbitrary Python through the MCP server is off by default and is enabled deliberately
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/fluent-cfd-skillDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The skill directs the agent to run Fluent and TUI commands through MCP, and calculations spend a license and compute time
- Arbitrary Python through the MCP server is off by default and is enabled deliberately
This is third-party code. Review the repository files before installing.
What it does
The skill sets the order of work for an agent doing CFD in Ansys Fluent. Fluent documentation stays the source of facts, the `pyfluent` MCP server is the execution layer, and the skill decides the order of operations and what to check. Fluent itself is not bundled. The skill leads from problem framing to result validation. It helps plan a case setup before touching the solver, review boundary conditions, mesh quality and physical assumptions, and choose the solver, turbulence model, wall treatment and convergence criteria. Diagnosis covers residual stalls, reversed flow, divergence and launch and license failures. In PyFluent MCP mode it describes a safe loop: launch, read case, TUI commands, iterate, save, close. Arbitrary Python is off by default until enabled in the local MCP server. `PROMPTS.md` holds templates for a full review, a short daily check and convergence diagnosis.
Who it is for. CFD engineers and researchers who run Ansys Fluent calculations with Codex and the PyFluent MCP server.
Good fit when
- A setup review is needed before a run: mesh, boundary conditions, physical assumptions
- Residuals stalled and you must tell poor convergence from an unphysical result
- You must pick a turbulence model and wall treatment for a specific case
- The agent runs the calculation through PyFluent MCP and needs a safe order for launch, edit and save
Not a fit when
- You have no Ansys Fluent license or installed solver: the skill does not bundle it
- You need a reference on general fluid dynamics theory: the skill points to Fluent documentation
- The calculation runs in another system such as OpenFOAM or COMSOL
Example request
Review the Fluent calculation setup in the case file and explain why the turbulence residuals are flatLimitations
- Ansys Fluent with a license and, for automation, a local `pyfluent` MCP server are required, and the skill installs neither.
- The skill is a judgment layer for research and does not replace the Fluent manuals, and its conservative defaults suit research use.
- Installation is described for Windows through `Copy-Item` into `.codex/skills`, with no ready commands for other systems in the README.
- The last commit dates from late September 2026, so check Fluent versions against the documentation.
How to disable. Delete the `fluent-cfd` skill folder from `~/.codex/skills/` (on Windows `%USERPROFILE%\.codex\skills\fluent-cfd`).
FAQ
How do I install the skill?
Copy the repository folder into the Codex skills directory, for example with `Copy-Item -Recurse . "$env:USERPROFILE\.codex\skills\fluent-cfd"`. After that Codex triggers the skill when a task mentions Fluent or PyFluent.
Is Fluent needed for the skill?
Yes, the skill does not bundle Fluent. It describes the order of work and checks, and your installed Fluent runs the calculation, for example through PyFluent MCP.
What does the skill check for convergence?
Residuals, monitors, conservation and physical plausibility together and not residuals alone. A separate recovery reference covers stalls and divergence.
Does the agent run arbitrary Python?
Not by default. It is enabled only explicitly in the local MCP server, which is why the skill gives a safe, auditable order of commands.
Are there ready prompt templates?
Yes, in `PROMPTS.md` and the Chinese `PROMPTS.zh-CN.md`: a full CFD review, a short daily check, a PyFluent MCP run and convergence diagnosis.
Discussion
No questions or reviews yet.
Related
SkillsA skill for Codex and Claude Code that turns ideas and bugs into working code through a writer, tester and reviewers
A library of agent skills for CAD, CAE and CAM: generate and edit models, export STEP, STL, 3MF, robot description files and mesh slicing to g-code
61 small skills for Codex across 13 areas: codebase orientation, tests, migrations, security, CI and pull request review
SolidWorks Automation Skill
solidworks-automation-skill
A skill and an MCP toolkit for automating desktop CAD (SolidWorks, AutoCAD): the agent calls the real program and checks the result. Windows, README in Chinese
An Abaqus modeling assistant for Codex: environment check, a step-by-step model, solve, ODB reading and a Chinese report
Python SDK for parametric CAD with an agent skill: code-built models, STEP, STL and MJCF export, edits rerun only dependent cells.
README in short
The README calls the skill a judgment layer for CFD in Fluent that does not replace the manuals. It shows highlights, the repository layout, use cases, prompt templates and installation by copying into the Codex skills directory. A Chinese version exists.
README badge
Original SKILL.md
--- name: fluent-cfd description: Guide Ansys Fluent and PyFluent CFD simulation workflows with MCP tool-use discipline. Use when Codex works on Fluent, PyFluent, Fluent TUI, CFD or fluid simulation setup, case/data files, mesh quality, boundary conditions, turbulence models, residual convergence, UDFs, post-processing, solver diagnostics, or validation of Ansys Fluent results. --- # Fluent CFD Use this skill as the workflow and judgment layer for Ansys Fluent work. Keep Fluent documentation as the fact source, use the local `pyfluent` MCP server as the execution layer, and use this skill to decide the order of operations, checks, and validation criteria. ## Operating Rules - Define the physics before touching the solver: objective quantity, geometry, units, fluid, regime, steady/transient behavior, compressibility, heat transfer, multiphase, rotating zones, and expected validation evidence. - Do not treat residual decrease alone as convergence. Require monitor stabilization and conservation checks for the quantities that matter. - Do not skip mesh quality, units, boundary-zone naming, boundary-condition consistency, and wall-resolution checks. - Do not change turbulence models without a physical reason or a stated validation purpose. - Prefer safe MCP tools and explicit Fluent TUI commands. Use `execute_python` only when `PYFLUENT_MCP_ENABLE_PYTHON=1` and the workflow is trusted. - Before starting Fluent or consuming a license, state that the action will launch Fluent and may occupy a license/session. - End live Fluent sessions with `exit_fluent` when the task is complete unless the user asks to keep the session open. ## Workflow 1. Frame the case: identify objective metrics, physics, known inputs, missing data, expected outputs, and acceptance criteria. 2. Choose solver and models: decide pressure-based vs density-based, steady vs transient, energy, multiphase, turbulence, wall treatment, and reference values. 3. Check the mesh and setup basis: units, cell/face zones, boundary names, quality metrics, non-orthogonality/skewness, and y+ target. 4. Set or review materials, boundary conditions, operating/reference conditions, numerics, initialization, and monitors. 5. Run a smoke test first: a short iteration/time-step run to catch setup errors before a full solve. 6. Iterate or advance time while monitoring residuals, integral balances, and objective quantities. 7. Save case/data and report validation status, remaining risks, and next checks. ## Reference Loading Load only the reference files needed for the current task: - `references/workflow.md`: complete Fluent workflow and task routing. - `references/solver-selection.md`: model and solver selection rules. - `references/boundary-conditions.md`: inlet, outlet, wall, symmetry, periodic, and rotating-zone boundary checks. - `references/turbulence-and-wall-treatment.md`: turbulence model, wall treatment, y+, and mesh implications. - `references/numerics-and-convergence.md`: discretization, initialization, monitors, residuals, and convergence. - `references/pyfluent-mcp.md`: local `pyfluent` MCP tools, call order, safety limits, and failure handling. - `references/validation-checklist.md`: conservation, mesh independence, physical plausibility, and post-processing checks. - `references/error-recovery.md`: divergence, reversed flow, negative volume, CFL, residual stalls, and launch/license failures. ## MCP Tool Discipline When a task requires real Fluent execution through the local MCP server: