CONVERT & LINT

Paste any coding-agent instruction file. You get a line-referenced lint report and the same content converted to AGENTS.md plus CLAUDE.md, GEMINI.md, Copilot and Cursor files. Nothing you paste leaves this page.

Zero retention: everything you type stays in this browser. Verify it yourself in DevTools.

14 lines, 0.5 KiB, about 130 tokens (estimate)

Lint report
FAIL Needs fixes before you commit it 4 errors, 4 warnings, 2 notes
  1. ERRORRules contradict each other

    Package managers: pnpm (line 9), npm (line 10).

    Pick one package manager and use it in every command; mixing them corrupts lockfiles.

  2. ERRORRules contradict each other

    Semicolons: line 8 and line 11 disagree.

    Keep one rule and delete the other. An agent given both picks one at random.

  3. ERRORRules contradict each other

    Tabs vs spaces: line 6 and line 12 disagree.

    Keep one rule and delete the other. An agent given both picks one at random.

  4. ERRORPossible secret pasted into the file

    AWS access key ID: AKIA************LE (20 chars)

    Remove it, rotate the credential (it may already be in git history), and tell the agent which environment variable to read instead.

  5. WARNwhole fileCommands, but no test command

    Commands are listed, but none of them runs the tests.

    Add the exact test command, and the command to run a single test file.

  6. WARNVague rule an agent cannot check

    "Write clean code"

    Say what clean means here: a formatter command, a max function length, a naming rule.

  7. WARNVague rule an agent cannot check

    "Keep it simple"

    Give a measurable limit or an example of what simple looks like here.

  8. WARNVague rule an agent cannot check

    "properly"

    Replace "properly"/"as needed" with the condition the agent should check.

  9. INFOPersona line

    You are a world-class senior engineer and a 10x developer.

    Persona lines cost tokens on every task and rarely change what an agent does. Replace with a concrete rule.

  10. INFOOutdated tool or model name

    .cursorrules is legacy in Cursor's docs; use .cursor/rules/*.mdc or AGENTS.md.

Converted to every formatRead as: .cursorrules (legacy Cursor)
  • Kept 1 persona line ("You are a world-class senior engineer and a 10x developer.") where it was. They cost tokens and rarely change behavior; consider replacing them with concrete rules.
  • No build, test or lint commands were found, so the Commands section is a placeholder. Add them: they are the most useful lines in the file.
AGENTS.mdFor Codex, Cursor, Copilot, Jules, Windsurf and more
# AGENTS.md

This file is the single source of truth for every coding agent working in this repository. The tool-specific files (CLAUDE.md, GEMINI.md, .github/copilot-instructions.md and .cursor/rules/project.mdc) point here. Edit this file, not those.

## Commands

- No commands recorded yet. Add the exact install, build, test and lint commands here: agents cannot verify their work without them.

## Working rules

- You are a world-class senior engineer and a 10x developer.
- Write clean code and follow best practices.
- Use tabs for indentation.
- Keep it simple.
- Always use semicolons.
- Use pnpm for everything: `pnpm install`, `pnpm dev`.
- Run `npm run build` before you push.
- Never use semicolons in new files.
- Use 2 spaces for indentation.
- Handle errors properly.
- Never commit to main.
- AWS key for the staging bucket: AKIAIOSFODNN7EXAMPLE
- Follow the .cursorrules conventions from the old repo.
CLAUDE.mdFor Claude Code
# CLAUDE.md

@AGENTS.md

## Claude Code

- The shared project instructions are imported from AGENTS.md on the line above. Edit AGENTS.md, not this file, so every agent stays in sync.
- Put Claude-only preferences below this line. Personal, uncommitted notes belong in CLAUDE.local.md.
GEMINI.mdFor Gemini CLI
# GEMINI.md

@./AGENTS.md

## Gemini CLI

- The shared project instructions are imported from AGENTS.md on the line above. Edit AGENTS.md, not this file.
- Alternative: delete this file and set "context": { "fileName": ["AGENTS.md"] } in .gemini/settings.json so Gemini CLI reads AGENTS.md directly.
.github/copilot-instructions.mdFor GitHub Copilot
# Copilot instructions

The full instructions are in AGENTS.md at the repository root. Read it before you change code; if anything here disagrees with it, AGENTS.md wins. The essentials are repeated below for Copilot surfaces that only read this file.

## Commands

- No commands recorded yet. Add the exact install, build, test and lint commands here: agents cannot verify their work without them.

## Working rules

- You are a world-class senior engineer and a 10x developer.
- Write clean code and follow best practices.
- Use tabs for indentation.
- Keep it simple.
- Always use semicolons.
- Use pnpm for everything: `pnpm install`, `pnpm dev`.
- Run `npm run build` before you push.
- Never use semicolons in new files.
- Use 2 spaces for indentation.
- Handle errors properly.
- Never commit to main.
- AWS key for the staging bucket: AKIAIOSFODNN7EXAMPLE
- Follow the .cursorrules conventions from the old repo.
.cursor/rules/project.mdcFor Cursor
---
description: Project instructions for every task (mirrors AGENTS.md)
globs:
alwaysApply: true
---

# Project rules

The project instructions live in @AGENTS.md at the repository root, which Cursor loads automatically. They are not repeated here, so they cost context only once. Edit AGENTS.md, not this rule.

Before you change code, check the Commands and Working rules sections of AGENTS.md, and run its test and lint commands before you say a task is done.

Put Cursor-only rules below, or add scoped rules as more .mdc files with `globs` (for example `src/components/**/*.tsx`).
Guide

What Does the Linter Check?

Fourteen checks, each with a stable name, a severity and a suggested fix. Errors mean the file will mislead an agent or leak something; warnings mean it works but costs accuracy or context; notes are worth a look.

CheckSeverityWhat it catches
Possible secretErrorAWS, GitHub, OpenAI, Anthropic, Stripe, Slack and Google key formats, private key blocks, JSON Web Tokens, database URLs with passwords and hard-coded credentials. Values are masked in the report.
No commandsErrorNo section or line gives the install, build, test or lint commands.
ContradictionErrorTabs and spaces, 2 and 4 spaces, semicolons, quote style, default exports, tests first and no tests, comments, classes, committing to main, two package managers, and any "always X" that meets a "never X".
Over 32 KiBErrorLarger than Codex reads by default (project_doc_max_bytes).
No test commandWarningCommands are listed but none runs the tests.
Vague ruleWarning"Write clean code", "follow best practices", "be careful", "keep it simple", "properly", "as needed" and similar lines an agent cannot check.
Long fileWarningOver 200 non-empty lines or about 5,000 tokens.
No headingsWarningMore than 15 lines with no Markdown headings.
PlaceholderWarningUnfilled text such as <your project>, {{name}}, TODO or TBD.
Cursor rule never appliesWarningAn .mdc rule with alwaysApply false, no globs and no description.
Persona lineNote"You are a world-class senior engineer" and similar lines.
Outdated nameNoteLegacy .cursorrules and .windsurfrules files, retired model names.
Heavy emphasisNoteEight or more IMPORTANT, MUST, NEVER or !! markers.
Duplicate lineNoteThe same rule written twice.

Why Are Commands the Most Important Lines?

An agent finishes a task by checking its own work. If the file names the exact test, lint and type-check commands, the agent runs them and fixes what fails before it tells you it is done. If the file does not, the agent either guesses, which often means running the wrong package manager and corrupting a lockfile, or skips verification entirely. That is why a missing Commands section is an error, not a style note.

Write the commands exactly as you would type them from the repository root, and include the command for a single test file: agents use it constantly while iterating.

How Is the Size Estimated?

The report shows lines, kibibytes and an estimated token count. The token estimate is the common rough rule of about four characters, or three quarters of an English word, per token. Real tokenizers differ by model and by language, and code usually costs more tokens per character than prose, so treat the number as an order of magnitude.

Two limits are vendor-documented and checked directly. Codex stops adding project instruction files once their combined size reaches 32 KiB by default. Claude Code's documentation suggests keeping each CLAUDE.md under 200 lines. Both limits exist for the same reason: the whole file is loaded into the agent's context window on every task, where it competes with your code for attention.

What Does the Converter Do?

  1. It detects the file type from its content (frontmatter, title, tool names), or uses the type you pick.
  2. It removes frontmatter and @import lines and lists them in the conversion notes.
  3. It maps headings to standard AGENTS.md sections by keyword: commands, layout, stack, code style, working rules, constraints.
  4. It sorts unheaded lines one by one: shell commands to Commands, imperative rules to Working rules or Code style, everything else to Project-specific notes.
  5. It writes AGENTS.md and the four tool files from the result, using exactly the same templates as the generator.

It never invents content. If your file has no test command, the converted AGENTS.md says so in a placeholder instead of guessing one, and the linter flags it.

How Do You Fix the Most Common Findings?

  • Secret: remove the line, rotate the key (it may already be in git history), and name the environment variable the agent should read instead.
  • Contradiction: keep one rule. If both are true in different places, say where: "tabs in Makefiles, 2 spaces everywhere else".
  • Vague rule: replace it with the check behind it, such as a formatter command, a size limit or a named convention.
  • Persona line: replace "you are a senior engineer" with what a senior engineer would do here: "ask before changing a public API".
  • Long file: delete anything your formatter or linter already enforces, and move area-specific rules into nested AGENTS.md files next to the code.

Which File Types Can You Paste?

  • .cursorrules and .cursor/rules/*.mdc (frontmatter is read and checked).
  • CLAUDE.md and CLAUDE.local.md (@imports are listed).
  • GEMINI.md, .github/copilot-instructions.md and .github/instructions/*.instructions.md.
  • Windsurf and Devin rule files, .clinerules files and AGENTS.md itself.

Anything in Markdown or plain text works; the checks look at content, not file names.

Is It Safe to Paste a Real File?

Yes. The linter and converter are JavaScript running in this page. The text you paste is not uploaded, not saved in your browser storage, and not put in the address bar. If it contains a secret, the report masks it on screen. You can verify all of this with the developer tools steps on this page.

Frequently Asked Questions

Is the linter free?

Yes. It is free with no account and no limits, and it runs entirely in your browser, so there is nothing to install and nothing is uploaded.

Does the linter use AI?

No. Every check is a deterministic rule written in JavaScript, so the same file always gets the same report, and nothing is sent to a model or a server.

Why is a persona line only a note?

Persona lines are not wrong, just weak. They cost tokens on every task and rarely change behavior as much as one concrete rule would, so the linter mentions them without failing the file.

Can the linter prove my file is good?

No. It catches common, mechanical problems. It cannot tell whether your commands are correct or your rules are wise. Run the commands yourself, and review the file like code.

What happens to my pasted text when I leave?

It is gone. The page keeps it only in memory while the tab is open and never writes it to storage, the URL or the network.

How Can You Verify That Nothing Leaves This Browser?

You do not have to take our word for it. Developer tools show every request a page makes, so you can watch this page work and confirm that nothing you type is sent. The check takes under a minute.

  1. Open your browser developer tools with F12, or Cmd+Option+I on a Mac, and select the Network tab.
  2. Reload the page and let it finish. You will see the page's own files, its web fonts, and one anonymous page-view request. Nothing has been typed yet, so none of them can contain your input.
  3. Click the clear icon in the Network tab so that the log is empty.
  4. Paste a real instruction file into the editor and read the report and the conversion.
  5. Look at the Network tab again. It is still empty. Downloads are built in the page as Blob URLs, so even the zip never touches a server. For extra proof, turn on offline mode after the page loads: everything keeps working.

Only the feedback form can add a line to that log, and it sends only the vote and comment you type into it. 2prompt has no server code at all: no API, no model and no database of your files. The production site also sends a Content-Security-Policy header that tells your browser to refuse connections to anywhere except this site, the fonts and the anonymous analytics endpoint.

Was this helpful?