English | 简体中文
A Codex skill for Ansys Fluent, PyFluent, and CFD simulation workflows.
This repository packages a research-oriented workflow skill for working with Fluent cases, solver setup, convergence diagnosis, and PyFluent MCP automation. It is designed as a judgment layer for agents: it does not replace the Fluent manuals, and it does not bundle a Fluent runtime.
- Fluent workflow discipline from problem framing to validation.
- PyFluent MCP tool-use guidance for safe, auditable Fluent automation.
- Convergence checks that combine residuals, monitors, conservation, and physical plausibility.
- Practical references for solver choice, boundary conditions, turbulence, numerics, validation, and error recovery.
- Conservative defaults for research use: no arbitrary Python execution unless explicitly enabled in the local MCP server.
.
+-- SKILL.md
+-- agents/
| +-- openai.yaml
+-- references/
+-- workflow.md
+-- solver-selection.md
+-- boundary-conditions.md
+-- turbulence-and-wall-treatment.md
+-- numerics-and-convergence.md
+-- pyfluent-mcp.md
+-- validation-checklist.md
+-- error-recovery.md
SKILL.md is intentionally compact. It routes the agent through the core CFD workflow and points to focused references only when needed.
- Plan a Fluent case setup before touching the solver.
- Review boundary conditions, mesh quality, and physical assumptions.
- Choose solver settings, turbulence models, wall treatment, and convergence criteria.
- Diagnose residual stalls, reversed flow, divergence, and launch/license failures.
- Guide PyFluent MCP workflows such as launch, read case, run TUI commands, iterate, save, and close.
Use the standard templates in PROMPTS.md for full CFD reviews, short daily checks, PyFluent MCP runs, and convergence diagnosis.
Copy this folder to your Codex skills directory:
Copy-Item -Recurse . "$env:USERPROFILE\.codex\skills\fluent-cfd"After installation, Codex can trigger the skill when a task mentions Fluent, PyFluent, Fluent TUI, CFD simulation setup, mesh quality, boundary conditions, turbulence models, residual convergence, UDFs, or post-processing.
Versions are managed with GitHub Releases and semantic version tags. Use the latest release for a stable skill snapshot, and use main for active edits.
- Residual decrease alone is not convergence.
- Mesh quality, units, boundary consistency, and wall-resolution checks should not be skipped.
- Turbulence model changes should have a physical reason or validation goal.
- The optional
execute_pythonMCP tool should stay disabled unless the local workflow is trusted. - This repository contains only the skill. A separate local PyFluent MCP server is required for live Fluent control.
The skill is designed to pair with a local pyfluent MCP server that exposes tools such as server_info, launch_fluent, read_case, run_tui, iterate, write_case_data, session_status, and exit_fluent.
That execution layer is intentionally separate from this repository so this skill remains lightweight, portable, and safe to publish.