Answers
Keep each Cursor rule under 500 lines and split the rest by file pattern
Published
Cursor’s own guidance is to keep each rule under 500 lines; past that, split it into several rules in .cursor/rules/ and make most of them apply only to matching files or when the agent judges them relevant, so an Always Apply rule holds only what every chat needs. Remember that rules never reach Tab or other non-Agent features, however short they are.
For Cursor. Vendor facts are from the pages linked below, read on November 18, 2026.
How to split a long rule
- Count:
wc -l .cursor/rules/*.mdc AGENTS.md. Anything near 500 lines is a candidate. - Keep one short Always Apply rule for commands and conventions every task needs.
- Move the rest into one file per subject with
globs:(Apply to Specific Files) or a cleardescription:(Apply Intelligently). - Move legacy
.cursorruleslines into.cursor/rules/or AGENTS.md. - Nested AGENTS.md files work too: Cursor reads them in subdirectories, and the more specific one takes precedence.
---
description: Payments webhooks
globs: services/payments/**
alwaysApply: false
---
Acknowledge with 200, then enqueue.The numbers
| Measure | Value | Source |
|---|---|---|
| Cursor’s recommended maximum per rule | 500 lines | Cursor rules docs |
| Rule types | 4: Always Apply, Apply Intelligently, Apply to Specific Files, Apply Manually | Cursor rules docs |
| AGENTS.md files longer than 500 lines (Cursor reads AGENTS.md) | 3.3% (weighted) | Harbor study, 18 Oct 2026 |
| AGENTS.md files longer than 200 lines | 17.8% (weighted) | Harbor study, 18 Oct 2026 |
Study figures are weighted estimates from 1,935 CLAUDE.md and 1,970 AGENTS.md files sampled from public GitHub repositories on 18 October 2026, with GitHub's own counts for the totals. Every figure is in the study's CSV, and the method is in the study.
What the docs say
- Cursor: Rules: Keep rules under 500 lines. Cursor supports AGENTS.md in the project root and subdirectories, and more specific instructions take precedence. Rules do not impact Cursor Tab or other AI features.
Where Harbor fits
Harbor connects Claude Code, Codex, Cursor and GitHub Copilot (the CLI and VS Code agent mode; the cloud agent is not supported yet) through hooks and an MCP server. In Cursor it adds the rules that apply when the session starts and on every prompt, picked for the task, so a rule about payments does not ride along on a CSS change. Each rule shows how often it was served and how often an answer cited it. A citation is evidence a rule was used, not proof it was obeyed.
Read next
- Cursor rules best practices
- Cursor Team Rules
- Check a Cursor rule for free
- Codex reads 32 KiB of AGENTS.md and cuts the rest: raise project_doc_max_bytes or trim
- Antigravity truncates each rule file at 24,000 bytes and caps always-on rules at 20,000 tokens
- Copilot code review no longer stops at 4,000 characters of instructions: the limit was removed in June 2026
- Every answer, by agent and problem
Questions
How long should a Cursor rule be?
Under 500 lines, per Cursor’s docs. Shorter is better for an Always Apply rule, since it goes into every Agent chat.
Do Cursor rules apply to Tab completions?
No. Cursor’s FAQ says rules do not impact Cursor Tab or other AI features; they apply to Agent chats.
Is .cursorrules still supported?
Cursor treats it as legacy. Move its lines into .cursor/rules/ or AGENTS.md, which Cursor reads at the root and in subdirectories.