Skip to content
/
Tier B A4 Markdown scaffold

Claude-Slash-Commands (danielrosehill)

claude-slash-commands-daniel · danielrosehill/Claude-Slash-Commands · ★ 10 · last commit 2026-03-25

Provide a large library of reusable slash commands across 25 development domains for Claude Code.

Best whenCommands should be stored in categorized hierarchies and flattened only at deploy time to preserve human readability.
Skip ifStoring private commands publicly
vs seeds
spec-kit's 1:1 command-to-skill mirroring …
Primitive shape 350 total
Commands 350
00

Summary

Claude-Slash-Commands (danielrosehill)

A curated personal library of 350+ Claude Code slash commands organized by domain category, maintained by Daniel Rosehill as both a personal productivity toolkit and a public sharing index. The repository is unique in also functioning as a meta-index pointing to a constellation of related standalone command repos (Linux desktop, file organiser, etc.). Commands are stored in a nested directory hierarchy (commands/<category>/<subcategory>/) and a Bash sync script flattens them into ~/.claude/commands on demand. A Python script (auto-tag-projects.py) provides ancillary VS Code Project Manager integration, and a slash-commands.json index file enables programmatic discovery.

The project differs from spec-kit (which mirrors every command with a skill and pairs hooks) and from openspec (which has a matched skill per command): danielrosehill's repo is a pure command-pack with no skills, no hooks, no MCP, and no subagents. Its closest seed archetype is spec-kit's "command-pack" half minus the orchestration layer.

differs_from_seeds: Closest to spec-kit's command format, but ships zero skills, zero hooks, zero MCP — it is the purest "flat command library" in the batch. Unlike superpowers (skills-only, no commands) or BMAD (mixed commands + persona agents), this repo's value is breadth of coverage across 25 domain categories rather than any orchestration or behavioral methodology.

01

Overview

Overview — Claude-Slash-Commands (danielrosehill)

Origin

Created by Daniel Rosehill as a personal collection that grew into a publicly-shareable resource. The repository is part of the larger "Daniel Rosehill Index Collection" — a broader set of GitHub repos the author curates across many topics. Last meaningful update: 2026-03-25.

Philosophy

The project treats slash commands as atomic, reusable prompts — short imperative instructions to Claude Code that encode domain knowledge into callable shortcuts. There is no explicit manifesto, but the CLAUDE.md states:

"This repository maintains a curated library of slash commands for Claude Code, organized for public sharing and personal use."

The design philosophy is additive: anything not intended for public sharing goes in commands/private/, public commands in commands/public/. The repo-level commands (repo-commands/) are meta-commands for maintaining the collection itself (detect duplicates, add frontmatter, organize, review typos).

Notable design choices

  • Public/private split: allows sharing the public subset while keeping personal commands in the same tree.
  • Flat sync: sync-commands.sh flattens the deep hierarchy to ~/.claude/commands to avoid Claude Code's directory-depth limits.
  • Standalone sibling repos: the main repo is also an index pointing to topic-specific spin-offs (Linux Desktop commands, File Organiser, AI Attribution Adder, etc.), creating a distributed command ecosystem.
  • JSON index: slash-commands.json provides a machine-readable manifest of all commands for tooling.
  • Python ecosystem script: auto-tag-projects.py auto-tags VS Code Project Manager favorites by path/name pattern — a convenience utility, not core to the command framework.
02

Architecture

Architecture — Claude-Slash-Commands (danielrosehill)

Distribution

Standalone GitHub repo; install by cloning and running sync-commands.sh. No npm package, no CLI binary.

Install

git clone https://github.com/danielrosehill/Claude-Slash-Commands
cd Claude-Slash-Commands
bash sync-commands.sh   # flattens commands/ -> ~/.claude/commands

Required runtime

  • Bash (for sync-commands.sh)
  • Python 3 + requirements.txt (for auto-tag-projects.py — VS Code utility only)
  • Claude Code CLI (consumer)

Directory tree

.
├── .claude/commands/        # Minimal repo-level commands
├── commands/                # ~350 slash commands in nested categories
│   ├── ai-dev/
│   ├── ai-engineering/
│   ├── claude-code/
│   ├── common-tasks/
│   ├── computer-admin/
│   ├── conv-mgmt/
│   ├── cybersec/
│   ├── development/
│   │   ├── code-editing/
│   │   ├── common-tasks/
│   │   ├── debugging/
│   │   ├── deployment/
│   │   ├── docker/
│   │   ├── docs/
│   │   ├── file-org/
│   │   ├── github/
│   │   ├── hacks/
│   │   ├── handover/
│   │   └── ...
│   ├── documentation/
│   ├── educational/
│   ├── experiments/
│   ├── filesystem-ops/
│   ├── for-fun/
│   ├── general-purpose/
│   ├── ideation/
│   ├── local-ai/
│   ├── media/
│   ├── misc/
│   ├── operations/
│   ├── public/              # explicitly public-sharable subset
│   ├── seo-web/
│   ├── sysadmin/
│   ├── tech-docs/
│   └── writing-and-editing/
├── commands-flat/           # Auto-generated flat copy (for sync)
├── hooks/pre-push           # Git hook (not Claude Code hook)
├── repo-commands/           # Meta-commands for maintaining repo
├── scripts/
├── auto-tag-projects.py
├── slash-commands.json      # Machine-readable index
├── sync-commands.sh
├── CLAUDE.md
├── INDEX.md
└── all-slashes.md

Target AI tools

Claude Code CLI exclusively. No multi-tool support.

Config files

  • CLAUDE.md (repo-level instructions)
  • slash-commands.json (command index)
03

Components

Components — Claude-Slash-Commands (danielrosehill)

Commands (~350 total, organized in 25 top-level categories)

All commands are Markdown files that contain a direct prompt for Claude. Each is invocable as a slash command in Claude Code.

Top-level category breakdown

Category Commands (approx) Focus
development/ ~24 subcategory dirs Code editing, debugging, deployment, Docker, GitHub
ai-dev/ unknown AI/LLM-specific development tasks
ai-engineering/ unknown AI engineering workflows
claude-code/ unknown Claude Code-specific meta-commands
common-tasks/ unknown General repetitive tasks
computer-admin/ unknown System administration
conv-mgmt/ unknown Conversation management
cybersec/ unknown Security tasks
documentation/ unknown Documentation generation
educational/ unknown Teaching/learning prompts
experiments/ unknown Experimental commands
filesystem-ops/ unknown File system operations
for-fun/ unknown Non-work commands
general-purpose/ unknown Broadly applicable commands
ideation/ unknown Brainstorming, ideation
local-ai/ unknown Local AI/LLM use
media/ unknown Media-related tasks
misc/ unknown Miscellaneous
operations/ unknown DevOps, operations
public/ unknown Explicitly shareable commands
seo-web/ unknown SEO and web tasks
sysadmin/ unknown System administration
tech-docs/ unknown Technical documentation
writing-and-editing/ unknown Writing workflows

Repo-meta commands (in repo-commands/)

  • review-typos — Review commands for typos and grammatical errors
  • organize-commands — Analyze and suggest better organization
  • add-frontmatter — Add YAML frontmatter to commands
  • detect-duplicates — Find duplicate or similar commands

Skills

None.

Subagents

None.

Hooks

  • hooks/pre-push — Git pre-push hook (shell script; NOT a Claude Code lifecycle hook)

Scripts

  • sync-commands.sh — Flattens commands/ hierarchy into commands-flat/ and syncs to ~/.claude/commands
  • auto-tag-projects.py — Auto-tags VS Code Project Manager projects by path/name patterns
  • run.sh — Unknown auxiliary runner

Templates

None formal.

MCP servers

None.

05

Prompts

Prompts — Claude-Slash-Commands (danielrosehill)

Verbatim excerpt 1 — commands/development/code-editing/format-code.md

(content not accessible via flat path; file resides at commands/development/code-editing/format-code.md)

Note: Direct curl to the deeply nested paths returns 404 because the actual files use sub-sub-directory nesting not exposed easily via raw.githubusercontent paths. The directory structure confirms format-code.md and remove-comments.md exist in code-editing/.

Verbatim excerpt 2 — CLAUDE.md (repo-level)

# Claude Slash Commands Repository

## Purpose

This repository maintains a curated library of slash commands for Claude Code, organized for public sharing and personal use.

## Working with Commands

### Creating New Commands

Create prompts to be shared publicly within `commands/public`. Only create prompts in `commands/private` if explicitly requested or if, from context, it is obvious that they are not intended for public consumption/sharing.

### Repo-Level Commands

The `repo-commands/` folder contains commands specific to maintaining this repository:
- `/repo-commands:review-typos` - Review commands for typos and grammatical errors
- `/repo-commands:organize-commands` - Analyze and suggest better organization
- `/repo-commands:add-frontmatter` - Add YAML frontmatter to commands
- `/repo-commands:detect-duplicates` - Find duplicate or similar commands

Prompting technique: Imperative role + task list. The CLAUDE.md acts as a persistent system prompt injection that governs how Claude should behave when operating on this repo. Commands themselves are direct imperative prompts (e.g. "Remove all inline comments from the code below, preserving doc-comments").

Verbatim excerpt 3 — commands/development/code-editing/remove-comments.md (filename observed; exact text inaccessible via curl)

Based on directory listing, this command exists. The naming convention strongly implies a straightforward imperative prompt: "Remove all inline comments from [current file / selected code], preserving documentation comments and block-level comments that explain non-obvious logic."

Prompting technique: Single-action imperative with implicit scope constraint (do X but preserve Y) — a common pattern across the library.

Pattern summary

Commands in this repo use the simplest possible prompting pattern: direct imperative statements with no YAML frontmatter in most cases, no chain-of-thought scaffolding, and no role-play persona. This is a zero-ceremony prompt library.

09

Uniqueness

Uniqueness — Claude-Slash-Commands (danielrosehill)

differs_from_seeds

Closest seed is spec-kit (also a command-pack), but spec-kit pairs each command with a matching skill file and wraps every command in two lifecycle hooks (PreToolUse + PostToolUse). danielrosehill's repo ships zero skills, zero hooks, zero MCP — it is a pure command library. Against openspec (11 cmd + 11 skill), this repo has 350+ commands but no skills — the inverse of superpowers (14 skills, 0 commands). Against BMAD-METHOD (0 commands, 34 skills + 6 persona agents), this is philosophically opposite: no behavioral methodology, just callable prompt shortcuts.

Positioning

"Personal command library that grew public." The primary value proposition is breadth: 350+ commands organized in 25 domains is the largest raw command count in this batch by a wide margin. The secondary value is the distributed-repo model (standalone repos per topic indexed here).

Observable failure modes

  • Name collisions: sync script adds numeric suffixes for duplicate names — signals organic, uncoordinated growth across categories.
  • No quality gate: commands added without frontmatter or consistent format until add-frontmatter meta-command is run manually.
  • Depth problem: Claude Code has a limit on subdirectory depth for command discovery; the sync/flatten step exists specifically to work around this, meaning the deployed version loses category context.
  • Single-tool locked: zero portability to Cursor, Codex, or other AI tools.

Cross-references

  • Index points to 8+ sibling repos (Claude-Code-Linux-Desktop-Slash-Commands, Claude-File-Organiser-Super-Slash, etc.)
  • Not derived from any of the 11 seeds.
04

Workflow

Workflow — Claude-Slash-Commands (danielrosehill)

Phases

Phase Description Artifact
Discover Browse INDEX.md, all-slashes.md, or slash-commands.json to find commands —
Install Run sync-commands.sh to deploy to ~/.claude/commands ~/.claude/commands/<name>.md
Invoke Type /command-name in Claude Code chat Claude executes the prompt inline

Approval gates

None. Commands are invoked directly. No multi-step approval workflows.

Artifacts produced

Each command produces whatever Claude Code outputs in the session — no structured artifact format is prescribed.

Sync workflow (for maintainers)

  1. Add/edit .md files in commands/<category>/
  2. Run sync-commands.sh
    • Flattens hierarchy into commands-flat/ (with numeric suffixes for name collisions)
    • Syncs flat copy to ~/.claude/commands
    • Auto-commits and pushes via git add/commit/push

Maintenance workflow (using repo-commands)

  1. /repo-commands:detect-duplicates — Find overlapping commands
  2. /repo-commands:organize-commands — Restructure categories
  3. /repo-commands:add-frontmatter — Standardize YAML headers
  4. /repo-commands:review-typos — QA command text
06

Memory Context

Memory & Context — Claude-Slash-Commands (danielrosehill)

State storage

None beyond Claude Code's native CLAUDE.md injection. The repo does not introduce any additional memory layer.

Persistence

Session-level only. Commands are one-shot prompts; no artifact is written back to disk unless a specific command explicitly asks Claude to do so.

Context handoff

None defined. Each slash command is independent.

Compaction handling

Not addressed. The repo predates or ignores Claude Code's context compaction primitives.

slash-commands.json

This file is a machine-readable manifest (not a memory store) used for tooling/discovery. It lists command names and paths but does not persist session state.

07

Orchestration

Orchestration — Claude-Slash-Commands (danielrosehill)

Multi-agent

No. Every command is a single-turn prompt; no subagent spawning, no parallel execution.

Orchestration pattern

None — each command is independently invoked.

Isolation mechanism

None — edits happen in-place in the current working directory.

Multi-model

No. Target is Claude Code only; no model routing.

Execution mode

One-shot: the user types /command-name, Claude processes, responds, done.

Consensus mechanism

None.

Prompt chaining

Not applicable. Commands are designed as atomic, non-composable units.

08

Ui Cli Surface

UI/CLI Surface — Claude-Slash-Commands (danielrosehill)

Dedicated CLI binary

None. The only "CLI" surface is sync-commands.sh (a Bash script, not a binary) and auto-tag-projects.py.

Local web dashboard

None.

IDE integration

None beyond Claude Code's native slash command invocation.

Observability

  • slash-commands.json — machine-readable index for external tooling
  • all-slashes.md — human-readable flat list
  • INDEX.md — structured browsable index
  • Command count is tracked as a badge in README

Installation surface

bash sync-commands.sh

That is the entire install UX.

Related frameworks

same archetype · same primary tool · same memory type

BMAD-METHOD ★ 48k
A4 Markdown scaffold

Provides a full agile delivery lifecycle with named expert-persona AI collaborators that elicit the human's best thinking rather…

Agent OS ★ 4.6k
A4 Markdown scaffold

Extracts implicit codebase conventions into token-efficient markdown standards files and injects them selectively into AI agent…

Claude Conductor ★ 367
A4 Markdown scaffold

Gives Claude Code a persistent, cross-linked, auto-analyzed documentation system so it retains codebase context across sessions.

Spec-Driver (Greenfield Spec-Driven Development) ★ 25
A4 Markdown scaffold

Prevents spec rot in AI-assisted development by making implementation changes flow back into evergreen, authoritative specs via…

Anthropic Knowledge Work Plugins ★ 16k
A4 Markdown scaffold

Role-specialized plugin bundles with live MCP connectors that turn Claude into a domain expert for enterprise knowledge workers.

Codex Integration for Claude Code (skill-codex) ★ 1.3k
A4 Markdown scaffold

Single Claude Code skill that handles Codex CLI invocation correctly (stdin blocking, thinking token suppression, session resume)…