A community-driven registry for Claude, Cursor, Windsurf, Cline & more. Not affiliated with Anthropic.
Are you the author? Sign in to claim
Copy and paste prompts for cursor agent. Free for anyone to use or add to. This repository provides a curated collection
⚠️ Archived — no longer maintained (as of June 2026). This repo is read-only. The configs still work and are public domain — copy what's useful. For current, living guidance see agents.md and the Cursor docs. See
SUNSET.mdfor why, andMIGRATION.mdfor the history.
Copy-paste configuration for getting the most out of Cursor and other
AI coding agents in 2026 — built around the primitives Cursor actually ships today:
AGENTS.md, scoped rules, subagents, Skills, and hooks.
This repo was rebuilt in 2026. It started in early 2025 as a pile of giant always-on
.mdcrules — an approach that's now an anti-pattern (it burns context and predates subagents/Skills/hooks entirely). If you used the old version, seeMIGRATION.md.
| Primitive | Lives in | Use it when… | Reads cross-tool? |
|---|---|---|---|
AGENTS.md | repo root (+ nested) | You want always-on project context: stack, conventions, commands. Start here. | ✅ Cursor, Codex, Copilot, Gemini CLI, Aider, Windsurf, Zed… |
Rules (.mdc) | .cursor/rules/ | You need scoped guidance — applies only to matching files (globs) or when the agent judges it relevant (description). | Cursor only |
| Subagents | .cursor/agents/ | A task needs its own isolated context or a specialized persona (review, verify, test) — possibly run in parallel. | Cursor (+ .claude/.codex dirs) |
| Skills | .cursor/skills/<name>/SKILL.md | You have an on-demand procedure that should load only when relevant (e.g. "write an ADR"), keeping context lean. | Cursor (+ compat dirs) |
| Hooks | .cursor/hooks.json + .cursor/hooks/ | You need deterministic, non-negotiable behavior around agent actions — format on save, block destructive commands. | Cursor |
Rule of thumb: AGENTS.md first. Add a scoped rule only when you hit a real scoping
need; reach for a subagent/Skill/hook when the job is delegation, on-demand knowledge, or
hard enforcement respectively.
cursor-agent-configs/
├── AGENTS.md # this repo's own agent context (a worked example)
├── templates/
│ ├── AGENTS.md # generic starter to copy into YOUR project
│ ├── nextjs/ # stack pack: Next.js + TS (AGENTS.md + scoped rule)
│ ├── python-uv/ # stack pack: Python + uv + ruff + pytest
│ └── go/ # stack pack: Go modules + table-driven tests
├── .cursor/
│ ├── rules/
│ │ ├── javascript.mdc # short, glob-scoped — defers style to ESLint/Prettier
│ │ └── typescript.mdc # short, glob-scoped — defers strictness to tsconfig
│ ├── agents/
│ │ ├── code-reviewer.md # read-only diff reviewer with a severity ladder
│ │ ├── verifier.md # confirms "done" work actually runs
│ │ └── test-author.md # writes regression-catching tests
│ ├── skills/
│ │ ├── debug/ # systematic debug loop (reproduce→isolate→…→verify)
│ │ ├── write-adr/ # ADR template, lifecycle, and when-to-write gate
│ │ └── init-agents-md/ # inspects a repo and drafts a tailored AGENTS.md
│ ├── hooks.json
│ └── hooks/
│ ├── format.sh # afterFileEdit: format the edited file
│ └── guard-destructive.sh # beforeShellExecution: block rm -rf /, force-push, …
├── examples/hooks/ # hooks cookbook: secret-guard, context-inject, test-on-save
├── docs/ # cheat sheet, cross-tool sync, troubleshooting
├── scripts/ # install.sh / install.ps1 / sync-agents.sh
└── MIGRATION.md
templates/: copy a ready AGENTS.md + matching
scoped rule for Next.js, Python (uv), or Go.scripts/sync-agents.sh generates
CLAUDE.md / .github/copilot-instructions.md / GEMINI.md from your AGENTS.md. See
docs/cross-tool.md.examples/hooks/: more ready-to-wire hooks (block
committing secrets, inject git context at session start, run the matching test on save)./init-agents-md — a Skill that inspects your repo and drafts an AGENTS.md for you.Copy what you want — nothing here requires a build step, Node, or npm.
git clone https://github.com/DVC2/cursor-agent-configs.git
# Option A: scripted (interactive, backs up anything it would overwrite)
cd your-project && /path/to/cursor-agent-configs/scripts/install.sh
# Option B: manual — copy only the pieces you want
cp -r cursor-agent-configs/.cursor/agents your-project/.cursor/
cp -r cursor-agent-configs/.cursor/skills your-project/.cursor/
cp cursor-agent-configs/templates/AGENTS.md your-project/AGENTS.md # then edit it
Then restart Cursor. Rules auto-attach by glob; Skills/subagents load on demand (invoke a
subagent with /code-reviewer, a Skill with /debug).
⚠️ Hooks run shell scripts on agent events. Read
.cursor/hooks/*.shbefore installing them, and only commit hooks your team trusts. They requirejq.
.ts/.js file — the matching rule attaches in the agent's context panel./ in Agent and confirm code-reviewer, debug, etc. appear.jq . .cursor/hooks.json should parse; try a blocked command to see the guard fire.Subagents and Skills require Cursor 2.4+ (nested subagents: 2.5+). AGENTS.md and
.cursor/rules work broadly. The legacy single-file .cursorrules is deprecated — migrate to
AGENTS.md + .cursor/rules.
docs/quick-reference.md — frontmatter & format cheat sheet for every primitive.docs/troubleshooting.md — "my rule/Skill/hook isn't firing."CONTRIBUTING.md — how to add a config.The Unlicense — public domain. Copy freely.
A practical approach to managing multiple AI agents in Cursor through strict file-tree partitioning and domain boundarie
📄 Configuration files that enhance Cursor AI editor experience with custom rules and behaviors
神奇海螺 cursorrules 阿拉丁神燈想要甚麼許願就有