Fill in your project once. Get AGENTS.md, CLAUDE.md, a Cursor .mdc rule with frontmatter and a SKILL.md skeleton, each ready to copy, plus lint notes from the vendors' docs.
AGENTS.md is a plain Markdown file that tells coding agents how your project works, and it has no required fields: the agents.md site says to use any headings you like and that the closest AGENTS.md to the edited file wins. This page fills the same content into four files. AGENTS.md uses the example headings from agents.md: Setup commands, Code style, Testing instructions and PR instructions. CLAUDE.md uses Commands, Testing and Pull requests and commits. The Cursor rule is the CLAUDE.md text under a .mdc frontmatter block. The SKILL.md skeleton puts the same sections under Instructions and adds an Examples heading for you to fill in. With the default form, AGENTS.md comes out with a Stack list, five commands, and a Code style, Do, Don't, Testing instructions and PR instructions section. A section you leave empty is not written.
The rules box is one text area with headings, because that is how people already write these files. Start a section with a line such as ## Code style, ## Do, ## Don't, ## Testing or ## Pull requests, then list one rule per line. Dashes, asterisks, bullets and numbers at the start of a line are cleaned up. Lines before the first heading count as code style. A heading the tool does not know, for example ## Security, is kept as its own section after the known ones. Keep rules concrete: Claude Code's memory docs give examples such as use 2-space indentation and run npm test before committing, and they advise a CLAUDE.md of under 200 lines. The lint notes tell you when the generated CLAUDE.md goes past that.
A Cursor project rule is a .mdc file in .cursor/rules, and a plain .md file there is ignored. Its frontmatter has three fields. With alwaysApply true the rule is always included and globs and description are ignored. With alwaysApply false and a globs pattern it attaches when a matching file is in context. With only a description the agent decides, and with neither it applies only when you @-mention it. The four choices on this page write exactly those combinations. For a skill, Anthropic's docs require a name of at most 64 characters made of lowercase letters, numbers and hyphens, without XML tags and without the words anthropic or claude, and a description that is not empty, at most 1,024 characters, without XML tags, and says what the skill does and when to use it. The notes check each of these, and the file name is built from your project name.
The tool writes starting files from what you type. It does not read your repository, so the commands and rules are only as good as your input. If a repository has both AGENTS.md and CLAUDE.md, Claude Code reads only CLAUDE.md, so start CLAUDE.md with @AGENTS.md to share one file; the notes repeat that. Only one Cursor file pattern is supported, because the docs show a single pattern. Agents change how they read these files, so compare the output with each tool's current docs.
Anything you would tell a new teammate: setup commands such as npm install, code style, how to run the tests and how to format a pull request. The agents.md site says there are no required fields and any headings work. This generator writes Setup commands, Code style, Testing instructions and PR instructions.
Claude Code's memory docs advise under 200 lines per file, because longer files use more context and reduce adherence. The lint notes count the lines of the generated CLAUDE.md and warn above 200.
It starts with a frontmatter block between two --- lines holding description, globs and alwaysApply, followed by Markdown. For example alwaysApply: true applies the rule to every chat, and globs: src/components/**/*.tsx with alwaysApply: false attaches it to matching files. The file goes in .cursor/rules and must end in .mdc.
The name is at most 64 characters of lowercase letters, numbers and hyphens, with no XML tags and no anthropic or claude. The description is required, at most 1,024 characters, and should say what the skill does and when to use it. Claude Code also cuts the listing text at 1,536 characters.
You can. Claude Code reads only CLAUDE.md when both exist, so put @AGENTS.md on the first line of CLAUDE.md and add Claude-specific lines below it. Other tools read AGENTS.md directly.
No. It is processed in your browser, in this tab, and no server receives it. Your choices in the form fields are saved in this browser, the text in the rules box is not.