codex

Extracts contributions, methods, evidence, assumptions, and limitations from a single research paper into structured notes.

1|Updated Jun 1, 2026
One-click install
npx skills add https://github.com/kridaydave/My_Skills --skill codex-kridaydave
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: codex
Source: https://github.com/kridaydave/My_Skills/tree/main/codex
Command: npx skills add https://github.com/kridaydave/My_Skills --skill codex-kridaydave

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Reading a paper superficially leads to misreading its claims — mistaking what the abstract promises for what the results actually prove, or missing a hidden assumption that breaks your use of it. This Skill performs a disciplined close read of one document, separating what the paper claims from what its evidence supports. ## Core Features & Use Cases - Structured Deep Read: Extracts the core contribution, problem framing, method mechanism, decoded equations, evidence, stated and hidden assumptions, and limitations into a consistent note format. - Claim vs. Evidence Gap Analysis: Explicitly flags where the abstract or title overstates what the experiments or proofs demonstrate, and marks the reader's own inferences separately from the paper's text. - Reproducibility Audit: Identifies undefined hyperparameters, missing details, and underspecified steps you would need to resolve before implementing or citing the work. - Use Case: Before building on a paper, ask for a deep read — you get the one-sentence contribution, the decoded method, the load-bearing assumption in §3.2 that won't hold in your setting, and a list of what's missing to reproduce the results. ## Quick Start Ask the assistant to deep-read the attached paper and extract its contribution, method, evidence, assumptions, and limitations before you build on it.

Frequently Asked Questions about codex

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I deeply understand a research paper before building on it?▼

Request a structured deep read that extracts the one-sentence contribution, the actual method mechanism, decoded equations, supporting evidence, and stated plus hidden assumptions. This separates what the paper claims from what its results demonstrate before you rely on it.

How to check if a paper's evidence supports citing it for a claim?▼

Ask for a citation check on the specific claim. The analysis compares the claim against the experiments and data actually used, flagging cases where the paper shows a narrower result — such as one synthetic dataset — than the general statement you want to cite.

Can this summarize multiple papers or survey a research field?▼

No, it is scoped to deep-reading one document at a time, not field surveys across many papers. For mapping a field across multiple papers, a survey-oriented skill such as Beacon is the designated complement.

What information is needed to reproduce a paper's method from scratch?▼

A reproduction-focused read lists undefined hyperparameters, missing implementation details, and glossed-over steps — for example unstated learning rates or tuning procedures. These gaps indicate where the paper's text alone is insufficient and released code is required.

Why does the analysis separate the paper's claims from the reader's critique?▼

Blending the paper's statements with personal interpretation causes misreading downstream. The format keeps faithful representation, labeled critique, and flagged inferences as distinct layers so you know exactly what is on the page versus inferred.