ieee-trans-skill

English | 简体中文

An open-source family of reusable Codex skills for evidence-grounded IEEE Transactions research workflows in electrical engineering.

The repository combines paper-to-experiment reasoning with publication-ready technical figure generation. Each skill can be used independently, while both follow the same provenance-first discipline: facts, derivations, design choices, and unknowns remain visibly distinct.

Skills

Skill	Use it for	Main results
ieee-trans-experiment-design	Deep-read a paper, audit its claims, and reverse-design reproducible, falsifiable experiments	Claim-evidence tables, technical audits, tiered experiment plans, experiment cards, missing-asset registers, and safety gates
make-ieee-trans-figures	Reconstruct and draw publication-ready electrical-engineering figures from source evidence and reference templates	Evidence-aware figure specifications, control-diagram templates, editable SVG, publication PDF, high-resolution PNG, and QA records

Current FRT showcase

The current positive showcase is the QA-passing FRT operator-chain detail. It keeps the FRT law explicit as state, error, gain, selector, scaling, saturation, current-circle, and min/max stages rather than hiding the logic in one opaque block.

Preview PNG

See the complete positive example bundle, including editable JSON, live-text SVG, vector PDF, PNG, and QA record.

Archived layout failure case

The former FRT dual-axis showcase is retained as an EXPECTED_FAIL regression fixture.

It is not a publication-ready or reusable layout reference. The repaired Skill now rejects its off-center and asymmetric ports, sub-arrowhead block gaps, avoidable routing, detached labels, and marker/bridge crowding at final physical size.

The original PNG and editable source remain frozen for regression and provenance. They must not be used as the positive showcase; the separately reviewed replacement above is the current approved example.

Shared evidence rules

Both skills use explicit provenance labels:

Label	Meaning
SOURCE_FACT	Directly supported by the supplied paper, model, data, or specification
DERIVED	Calculated or logically inferred from stated source facts
RECOMMENDATION	A proposed engineering or presentation choice
UNKNOWN	Not supported strongly enough to assert

Experiment plans remain NOT_EXECUTED until actual outputs and logs are verified. Missing source material is recorded rather than silently filled in. Selected operating points are not expanded into full-domain claims, and unverified numerical values are not presented as measured facts.

Setup

Use Python 3.10 or newer and install the rendering dependencies:

python -m pip install -r requirements.txt

The figure skill validates the actual font files used in an export. Install Times New Roman for Latin text and SimSun for Chinese text before rendering a figure that requires those fonts.

Control-diagram templates

make-ieee-trans-figures includes three built-in control-diagram templates:

• dual-axis-loop
• nested-observer-expanded
• reduced-observer-equivalent

These are legacy topology scaffolds, not layout-approved publication figures. They intentionally remain available for semantic reuse, but they do not carry an enforced layout contract and cannot receive overall PASS. After copying one, redesign its geometry against the physical layout gate; do not treat its old coordinates or waypoints as a visual template.

Blocks carry concise functional names by default. Equations appear only when the mathematical operator is essential to understanding the structure, such as a transfer function, integrator, resonant term, or reduced-order equivalent. Controller gains, ratings, and design notes stay in the specification or supporting documentation unless they must be visible for the figure's argument.

List and copy the templates:

python skills/make-ieee-trans-figures/scripts/scaffold_control_template.py --list
python skills/make-ieee-trans-figures/scripts/scaffold_control_template.py \
  --template dual-axis-loop \
  --output control-diagram.json

After reviewing the generated JSON specification, validate and render it:

python skills/make-ieee-trans-figures/scripts/render_control_diagram.py \
  control-diagram.json \
  --output-dir build \
  --basename control-diagram

The rendering workflow keeps the source specification editable and produces SVG, PDF, and PNG exports together with machine-readable QA results.

Run the contract and renderer test suite with:

python -X utf8 -m unittest tests/test_make_ieee_trans_figures.py

Experiment-design workflow

Use ieee-trans-experiment-design when you want to:

• infer experiments from a paper;
• reconstruct or reproduce an experiment design;
• validate or falsify a paper claim;
• design analytical, simulation, SIL, HIL, prototype, or extension tests;
• identify experimental gaps and missing assets;
• challenge claims such as full-range ZVS or global optimality.

The minimum input is a paper PDF. Supplementary material, source code, models, raw data, and component or instrument datasheets strengthen the reconstruction. Hazardous hardware work remains behind local laboratory procedures, verified protection, qualified supervision, and expert approval.

Repository layout

skills/
├── ieee-trans-experiment-design/
│   ├── SKILL.md
│   ├── agents/
│   ├── references/
│   └── examples/
└── make-ieee-trans-figures/
    ├── SKILL.md
    ├── agents/
    ├── assets/
    │   └── templates/
    ├── references/
    ├── scripts/
    └── examples/
        └── frt-dual-axis/

Publication and safety boundary

This repository supports research design, technical auditing, and figure production. It does not authorize hazardous laboratory operations. It does not redistribute restricted papers, paid PDFs, confidential reviews, unpublished data, or publisher-owned figures.

This project is not affiliated with, endorsed by, or sponsored by IEEE and does not guarantee acceptance by any journal.

License

Released under the MIT License.