Computer Architecture Research Skill

An argument-first Codex skill for shaping, reviewing, and polishing computer architecture research papers.

This skill is designed for work targeting venues such as ISCA, MICRO, HPCA, ASPLOS, PACT, CGO, SC, and MLSys. It helps keep the paper's claim, evidence, evaluation, figures, rebuttal, and artifact aligned before polishing the prose.

What It Helps With

• Review or rewrite abstracts, introductions, results, discussions, titles, conclusions, and rebuttals
• Design stronger evaluation sections for architecture papers
• Check workload coverage, baseline fairness, simulator fidelity, overheads, ablations, and sensitivity studies
• Improve architecture figures such as speedup charts, scalability curves, breakdowns, and schematics
• Prepare artifact and reproducibility material
• Surface reviewer risks before submission or rebuttal

Core Workflow

The skill follows a research-first workflow:

1. Classify the paper type
2. State the one-sentence core claim
3. Map the evidence chain
4. Check the evaluation contract
5. Identify likely reviewer risks
6. Then work on writing, figures, rebuttal, or artifacts

The central rule is simple: fix the research argument before fixing the prose.

Repository Contents

.
|-- SKILL.md
|-- agents/
|   `-- openai.yaml
`-- references/
    |-- artifact-reproducibility.md
    |-- evaluation-contract.md
    |-- figure-patterns.md
    |-- paper-types.md
    |-- reviewer-risk-checklist.md
    `-- writing-moves.md

Installation

Clone or copy this repository into your Codex skills directory:

git clone https://github.com/n0thingNoob/computer_architecture_research_skill.git

Then place the repository at:

~/.codex/skills/computer-architecture-research

or copy its contents into an existing skill folder with that name.

Example Prompts

• Help me polish this HPCA abstract.
• Is this MICRO evaluation section strong enough?
• Help me design the experiments for a GPU simulator paper.
• How should I revise this speedup figure for an architecture paper?
• A reviewer says my baseline is unfair. Help me draft the rebuttal.
• Review this artifact appendix for reproducibility gaps.

Design Notes

This skill borrows the argument-first discipline of high-quality scientific writing workflows, but it is tuned for computer architecture research rather than generic academic polishing. It does not invent claims, results, references, simulator validation, or performance improvements. When the evidence is weak, it is supposed to say so.