mechanical-engineering-research

Mechanical Engineering Research Skill | Thermal-Fluid Research Workflow Plugin

English | 简体中文 | 繁體中文

A domain-rigor layer and modular skill suite for thermal-fluid mechanical engineering research with AI agents.

Generic research agents can summarize papers and draft prose. This plugin helps them do the harder mechanical-engineering work: check heat-transfer and fluid-flow assumptions, catch invalid correlation use, question CFD validation, protect uncertainty analysis, explain mechanisms, and turn evidence into decision-ready research artifacts.

Use it when a thermal-fluid answer needs to be physically defensible, not just well written.

The suite keeps task-specific instructions small and discoverable while retaining one coordinator for work that crosses boundaries:

If this helps your research workflow, please star the repo so other mechanical-engineering researchers can find it.

Want the fastest path? Start with QUICKSTART.md.

The plugin assists reasoning and workflow; it does not replace the investigator's scientific judgment, source verification, authorship responsibility, or data-rights obligations.

Two-Minute Demo

Paste a prompt like this after installing the skill:

Use the mechanical-engineering-research skill to review this claim:

"The CFD model proves that the new microchannel heat sink is optimal because
the average Nusselt number is 40% higher than the baseline. The simulation used
k-epsilon turbulence, a coarse wall mesh, constant water properties, and three
flow rates between Re = 350 and 900. Pressure drop is not discussed."

Expected style of response:

Bottom line: The claim is overconfident. The result may indicate better heat
transfer in the tested cases, but it does not prove optimality.

Key issues:
- Re = 350-900 may be laminar or transitional for the stated geometry, so a
  k-epsilon turbulence model needs justification.
- A coarse wall mesh weakens confidence in wall heat flux and Nusselt number.
- Constant water properties may be acceptable only over a small temperature
  range; the temperature span must be stated.
- A 40% Nusselt increase is incomplete without pressure drop, pumping power,
  and thermal resistance.
- "Optimal" requires a design space, objective function, constraints, and
  uncertainty or grid-independence evidence.

Next checks: compare against laminar correlations or a conjugate heat-transfer
baseline, report y+ or wall treatment, run mesh independence, include pressure
drop and pumping power, and rewrite the claim as evidence from a limited CFD
study rather than proof of global optimality.

Workflow

%%{init: {"theme": "base", "themeVariables": {"fontFamily": "Inter, Segoe UI, Arial, sans-serif", "primaryColor": "#ecfeff", "primaryTextColor": "#0f172a", "primaryBorderColor": "#0f766e", "lineColor": "#475569", "secondaryColor": "#fff7ed", "tertiaryColor": "#f8fafc", "clusterBkg": "#ffffff", "clusterBorder": "#cbd5e1", "edgeLabelBackground": "#ffffff"}}}%%
flowchart TB
    A["Research Request"]:::input
    B["Plugin Router"]:::core

    C["Academic Scaffold"]:::scaffold
    D["ME Judgment Layer"]:::core

    A --> B
    C -. "process" .-> B
    B --> D

    D --> E{"Mode"}:::gate

    E --> F1["Literature Map"]:::lane
    E --> F2["Analysis + DOE"]:::lane
    E --> F3["CFD + Tests"]:::lane
    E --> F4["Writing + Proposals"]:::lane
    E --> F5["Code + AI/ML"]:::lane
    E --> F6["Slides + IP"]:::lane

    F1 --> G
    F2 --> G
    F3 --> G
    F4 --> G
    F5 --> G
    F6 --> G

    G["Rigor Gate"]:::gate
    H["Decision Output"]:::output
    I["Reusable Artifact"]:::artifact

    G --> H --> I

    classDef input fill:#eff6ff,stroke:#2563eb,stroke-width:2px,color:#0f172a;
    classDef core fill:#ccfbf1,stroke:#0f766e,stroke-width:3px,color:#0f172a;
    classDef scaffold fill:#f8fafc,stroke:#64748b,stroke-width:2px,color:#0f172a;
    classDef lane fill:#fff7ed,stroke:#f97316,stroke-width:2px,color:#0f172a;
    classDef gate fill:#fef3c7,stroke:#d97706,stroke-width:3px,color:#0f172a;
    classDef output fill:#ecfdf5,stroke:#16a34a,stroke-width:3px,color:#0f172a;
    classDef artifact fill:#f5f3ff,stroke:#7c3aed,stroke-width:2px,color:#0f172a;

Editable Mermaid source: assets/workflow.mmd.

What It Catches

• Correlations used outside their Reynolds, Prandtl, geometry, roughness, orientation, or phase-change validity range.
• CFD claims without mesh independence, wall treatment, convergence, boundary-condition, property-model, or validation evidence.
• Experiment plans missing sensor calibration, uncertainty propagation, repeatability, heat-loss correction, or flow-development checks.
• AI/ML workflows with leakage across videos, surfaces, experiments, geometries, pressures, or simulation families.
• Literature reviews that list papers chronologically instead of synthesizing mechanisms, methods, gaps, and benchmark evidence.
• Proposal sections that describe ambitious methods but do not connect barrier, capability, validation, metrics, risk, and impact.
• Results di