THE LANGUAGE PLAYGROUND

Small edits.
Sharper instructions.

Take an instruction apart. Change a few words.
See what the linter notices.

Live scans need JavaScript. You can still try the CLI example or read the playground guide.

Tool descriptions

Two tools. Which one?

Scan the original. Then make one description more specific and scan again.

01 / 05

Your input

Format

Original example437 / 20,000

⌘ / Ctrl + Enter to scan

Local scan · input stays in this tab. First scan downloads the engine.

What the linter sees

Ready for your first scan.

Would you know which tool to pick?

Run the example. Look for the tool-description finding, then try making the distinction clearer.

EDIT → SCAN → EXPLORE

LintLang 0.8.3 · runs locallyWhat a scan can’t tell you ↗

There’s more than one way to write a good instruction. Use findings as clues, then test how your agent actually behaves. Share playground feedback ↗

TRY IT ON YOUR OWN CODEBASE

Now scan
your project.

Find instruction and configuration issues in the agent setup you actually use.

Scan your project

Recognizes AGENTS.md, CLAUDE.md, SKILL.md, and other supported agent files. See coverage and custom paths. Guides open in a new tab, keeping your draft here.

Open a terminal in your project root. Requires uv and Python 3.10+.

uvx --from lintlang==0.8.3 lintlang scan --discover .

The first run downloads LintLang. Scans run locally and report findings without changing files. Add a CI threshold later.

Reproduce this browser input in your terminal

2. From its saved folder, run:

Same engine version and exact input. --fail-on fail opts into exit code 1 for FAIL findings.

Small edits. Sharper instructions.