edvardgrishin27

Nativeprompt — AI skill for Claude Code

AI community

Rewrite your prompt into the native dialect of the model you're actually running (Claude Code / Codex) — by each vendor's OFFICIAL rules — and see which rule drove each edit.

How to install Nativeprompt

This entry records only its repository, not the path inside it, so there is no exact command to give. Open edvardgrishin27/nativeprompt and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Nativeprompt does

Rewrite your prompt into the native dialect of the model you're actually running (Claude Code / Codex) — by each vendor's OFFICIAL rules — and see which rule drove each edit. Self-updating. Zero deps.

Alternatives in AI

PromptEnhance — Rewrite the following prompt to be clearer, more specific, and more effective for an AI assistant 250 ★
Writing Style Skill — Writing style skill template with built-in auto-learning 200 ★
Amux — Bidirectional LLM API adapter with IR pattern 117 ★

README

nativeprompt

**Rewrite your prompt into the native dialect of the model you're actually running — using that vendor's official rules — and show you which rule drove each edit.**

![MIT](https://img.shields.io/badge/license-MIT-black) ![Python 3.9+](https://img.shields.io/badge/python-3.9%2B-blue) ![zero deps](https://img.shields.io/badge/dependencies-0-brightgreen) ![281 tests](https://img.shields.io/badge/tests-907%20passing-brightgreen) ![offline](https://img.shields.io/badge/core-deterministic%20%C2%B7%20offline-lightgrey)

**English** | [Русский](README.ru.md)

Scope: the **agentic CLIs** — Claude Code, Codex, Gemini CLI, Grok Build, Qwen Code and Kimi CLI. Not the API, not the web chat.

Coverage differs by vendor, and the tool says so out loud. Anthropic and OpenAI publish model-specific prompting rules, so for Claude Code and Codex you get both halves: what to change in the prompt *and* which command to launch with. Google, xAI, Alibaba and Moonshot publish no such rules — for Gemini CLI, Grok Build, Qwen Code and Kimi CLI you get the harness half only. Inventing rules and attributing them to a vendor would defeat the point: every finding here cites a source.

Why

The two vendors do not agree on what a good prompt looks like. Their own docs say so.

Same phrase in your prompt	Claude Code	Codex / GPT‑5.x
"think step by step"	acceptable scaffolding	remove it — reasoning models plan internally, and prescribed intermediate steps get in the way (reasoning guide)
"double-check yourself"	warned about on Opus 5 — the model already verifies, so the reminder buys over-verification, tokens and latency; the tool flags the phrase but does not cut it (Opus 5 prompting)	harmless, but usually redundant
"only report the important stuff"	rephrase — Opus 5