Promptshelf Slash Library · 50 commands · 6 categories
Slash Command Snippet Library · 50 ready-to-paste prompts

Paste any of 50 slash commands into your Claude Code setup.

Six workflow categories — debug, review, refactor, test, ship, docs — each with a tight set of battle-tested prompts. Click a category chip to filter, type in the search box to narrow further, copy any card's body into .claude/commands/<slug>.md. Your active filters + search are encoded in the URL — bookmark or share your curated view.

50 / 50 visible 6 / 6 categories on 50 commands · 6 categories · zero accounts
Bay 01

Filter & search.

Categories
Search
Showing all 50 commands.
No commands match these filters. Toggle a category on or clear the search.
Bay 02

How to install a slash command.

Each card's body is a complete prompt template. Copy it, paste into a file at .claude/commands/<slug>.md in your project (or ~/.claude/commands/ for global), and Claude Code will register the command automatically. Invoke it in any session by typing /<slug> — the template runs with the current file context attached.

For the full slash-commands reference (argument syntax, frontmatter, MCP integration), see our Slash Commands Guide. For packaged workflow kits, grab the Claude Code Starter Kit — it ships 10 skill templates and a hooks cookbook alongside these command snippets.

Bay 03

Compound these with our other tools.

Reference
Slash Commands Guide
Syntax, frontmatter, argument handling, and MCP-tool bindings for Claude Code slash commands.
Game
Copilot Config Bingo
Audit 25 AI-coding habits including "Reusable skill / slash templates" — see which ones you've shipped.
001 · Free ZIP
Claude Code Starter Kit
10 skill templates, 5 CLAUDE.md templates, hooks cookbook, settings.json profiles, power prompts.
Bay 04

FAQ.

Where do the 50 commands come from?
They're hand-written templates distilled from common Claude Code workflows — the same pattern our Starter Kit uses for its skill files. Each template is a complete prompt body (usually 5–15 lines) that expects a single action from Claude Code: inspect a diff, write a test, draft a commit, etc. None are LLM-generated boilerplate; each one is a prompt we or the kit authors would actually paste into our own projects.
How do I install one?
Copy the card's body. Create .claude/commands/<slug>.md inside your project (or ~/.claude/commands/<slug>.md for cross-project). Paste. Claude Code picks it up on the next session. Invoke with /<slug>. For advanced options — arguments, YAML frontmatter, allowed-tools — read our Slash Commands Guide.
Why these six categories?
Debug, review, refactor, test, ship, and docs cover the phases most programming sessions actually pass through. They match the skill-groups shipped in our Claude Code Starter Kit and the pattern most teams discover on their own after a few months with Claude Code. Adding a seventh category is almost always a sign that one of these six is too coarse — most "new" categories collapse into refactor or review.
How does the share URL work?
Each category maps to one bit (debug=0, review=1, refactor=2, test=3, ship=4, docs=5). The six bits pack into a 2-character lowercase hex string — ?c=3f means all six on, ?c=01 means only debug. Add &q=<string> for search. ?c=3f with no q shows everything. The URL updates live as you filter; copy it, share it, bookmark it.
Can I adapt these for Cursor or Copilot?
Yes. The prompt bodies are tool-agnostic. For Cursor, paste into .cursor/rules/<slug>.md or load via Agent-mode instructions — see our Cursor Rules Starter Kit. For Copilot, save as a .prompt.md file and register via workspace settings — see our Copilot Customization Starter Kit. The only tool-specific detail in these templates is occasional reference to Claude Code's TodoWrite or Bash tools; swap those for your tool's equivalent when adapting.
Are these reviewed?
Every kit on this site has been through adversarial review (Codex GPT) and a smoke test where applicable. These 50 templates follow the same patterns shipped in our free kits — the /commit, /review, /debug, /ship skills in the kit's skill-templates/ directory were drafted with the same discipline. Nothing here promises behavior Claude Code doesn't deliver natively.
Is the page sending any data?
No. Filters, search, and share-URL encoding run entirely in the browser. No analytics, no telemetry, no third-party scripts beyond Google Fonts (CSS-only). Your filter state lives in the URL and optionally localStorage. Nothing leaves your device.
How do I tip the maker?
BTC: bc1qs04leape97ner4wqa98n94l9n0gv9aa84eg4ux (copy button in the tip jar below). No accounts, no signup, no middleman. If a command here saved you an hour, a few thousand sats is a fair thanks — but everything on Promptshelf stays free whether or not you tip.

Saved you an hour? Tip the maker in BTC — no account, no signup, just paste.

BTC bc1qs04leape97ner4wqa98n94l9n0gv9aa84eg4ux

Try one of our free tools

Hand-built single-file games, quizzes, and visualizers — no signup, no tracking, no cost.

Puzzle Prompt Injection Puzzle 10-level defender challenge — spot the injection before it runs. Game Copilot Config Bingo 25 AI-coding habits — how many have you adopted? Tool Prompt Token Visualizer Paste any prompt, see the token breakdown for GPT & Claude.