Skip to content
/

wshobson/commands Slash Command Collection

wshobson-commands · wshobson/commands · ★ 2.5k · last commit 2025-10-12

Primitive shape 57 total
Commands 57
00

Summary

wshobson/commands — Summary

wshobson/commands is a collection of 57 production-ready Claude Code slash commands (15 workflows + 42 tools) from the same author as wshobson/agents. The README explicitly positions this repo as superseded: "For a more modern approach, consider the Plugin Marketplace which provides similar functionality through a cleaner plugin architecture." The commands are organized into tools (single-purpose utilities) and workflows (multi-agent orchestration sequences that explicitly use the Task tool to spawn named subagents like backend-architect, frontend-developer, test-automator, deployment-engineer). Workflow commands such as feature-development describe a complete multi-agent pipeline in markdown, calling Task tool with subagent_type parameters and explicit prompt templates. The repo uses no skills, no hooks, no MCP — pure slash-command markdown files installed by cloning to ~/.claude. It differs from seeds primarily by being a pure command-only collection with explicit Task-tool orchestration instructions embedded in workflow markdown files, in contrast to superpowers (skills-only, no commands) and spec-kit (commands mirrored by skills).

01

Overview

wshobson/commands — Overview

Origin

Same author as wshobson/agents (Will Hobson). Older predecessor repo (2,482 stars, last push 2025-10-12). The README explicitly says:

"Note: This repository uses the slash commands pattern. For a more modern approach, consider the Plugin Marketplace which provides similar functionality through a cleaner plugin architecture."

Philosophy

Commands as first-class orchestration entry points. Workflows are multi-agent sequences where the command's markdown instructs Claude Code to use the Task tool with specific subagent types and prompt templates. Tools are single-purpose utilities with domain-specific knowledge embedded.

Organization

Two directories:

  • tools/ — 42 single-purpose commands (security scan, API scaffold, TDD steps, DB migrate, etc.)
  • workflows/ — 15 multi-agent orchestration commands (feature-development, full-review, incident-response, etc.)

Invocation

With directory prefixes:

/workflows:feature-development implement OAuth2 authentication
/tools:security-scan perform vulnerability assessment

Or flat (copy files to root):

/feature-development implement payment system

Status

Dormant (last push 2025-10-12). Superseded by wshobson/agents plugin marketplace.

02

Architecture

wshobson/commands — Architecture

Distribution

Clone to ~/.claude directly. No plugin marketplace, no adapter framework.

cd ~/.claude
git clone https://github.com/wshobson/commands.git

Directory Structure

tools/           # 42 single-purpose .md files
  security-scan.md
  api-scaffold.md
  tdd-red.md
  tdd-green.md
  tdd-refactor.md
  debug-trace.md
  db-migrate.md
  ...
workflows/       # 15 multi-agent orchestration .md files
  feature-development.md
  full-review.md
  security-hardening.md
  incident-response.md
  ...
examples/        # usage examples

Install

cd ~/.claude
git clone https://github.com/wshobson/commands.git

# Optional: flat install (no prefixes)
cp tools/*.md ~/.claude/
cp workflows/*.md ~/.claude/

Required Runtime

None beyond Claude Code.

Target AI Tools

Claude Code only.

No Build Step

Unlike wshobson/agents, there are no adapters, no Makefile, no generate scripts. Files are consumed directly.

03

Components

wshobson/commands — Components

Totals

Type Count
Workflow commands 15
Tool commands 42
Total commands 57
Skills 0
Subagents 0 (referenced by name in workflows; not shipped)
Hooks 0
MCP servers 0

Workflow Commands (15) — Multi-Agent Orchestration

Command Agents Coordinated
feature-development backend-architect, frontend-developer, test-automator, deployment-engineer
full-review Architecture, security, performance, quality reviewers
smart-fix Dynamic agent selection based on issue type
tdd-cycle Test writer, implementer, refactoring specialist
git-workflow Version control process automation
improve-agent Prompt engineering, performance tuning
legacy-modernize Architecture migration, dependency updates, pattern refactoring
multi-platform Web, mobile, desktop coordination
workflow-automate Build, test, deploy, monitor
full-stack-feature Backend API, frontend UI, mobile, database
security-hardening Threat modeling, vulnerability assessment, remediation
data-driven-feature Data science, feature engineering, model deployment
performance-optimization Profiling, caching, query optimization, load testing
incident-response Diagnostics, root cause analysis, hotfix deployment
ml-pipeline ML pipeline orchestration

Tool Commands (42) — Single-Purpose Utilities

AI and Machine Learning

  • ai-assistant — LLM integration, conversation management
  • ai-review — ML code review
  • langchain-agent — LangChain agent creation
  • prompt-optimize — Prompt engineering + cost optimization

Development Quality

  • tdd-red — Write failing test
  • tdd-green — Implement to pass
  • tdd-refactor — Refactor while green
  • code-migrate — Framework/version migration
  • refactor-clean — Clean refactoring
  • tech-debt — Technical debt identification

Security & Compliance

  • security-scan — OWASP vulnerability assessment
  • compliance-check — Regulatory compliance
  • deps-audit — Dependency vulnerability audit

Infrastructure

  • api-scaffold — REST endpoint scaffolding
  • api-mock — Mock API creation
  • db-migrate — Database migration management
  • k8s-manifest — Kubernetes manifest generation
  • deploy-checklist — Pre-deployment validation
  • docker-optimize — Docker optimization

Debugging & Analysis

  • debug-trace — Systematic debugging
  • error-analysis — Root cause analysis
  • error-trace — Error tracing
  • smart-debug — Intelligent debugging

Documentation

  • doc-generate — Documentation generation
  • code-explain — Code explanation

Context Management

  • context-save — Save session context
  • context-restore — Restore saved context

Testing

  • test-harness — Test framework setup
  • accessibility-audit — A11y compliance
  • data-validation — Data validation testing

Optimization

  • cost-optimize — Cloud cost optimization
  • monitor-setup — Monitoring setup

Collaboration

  • onboard — New team member onboarding
  • standup-notes — Daily standup generation
  • pr-enhance — PR description enhancement
  • issue — GitHub issue creation
05

Prompts

wshobson/commands — Prompt Excerpts

Excerpt 1: feature-development.md (workflows/)

Technique: Explicit Task-tool invocation with named subagent_type parameters — prompt chaining pattern where each agent receives prior agent's output as injected context.

---
model: claude-opus-4-1
---

Implement a new feature using specialized agents with explicit Task tool invocations:

[Extended thinking: This workflow orchestrates multiple specialized agents to implement a complete feature from design to deployment. Each agent receives context from previous agents to ensure coherent implementation.]

## Traditional Development Steps

1. **Backend Architecture Design**
   - Use Task tool with subagent_type="backend-architect" 
   - Prompt: "Design RESTful API and data model for: $ARGUMENTS. Include endpoint definitions, database schema, and service boundaries."
   - Save the API design and schema for next agents

2. **Frontend Implementation**
   - Use Task tool with subagent_type="frontend-developer"
   - Prompt: "Create UI components for: $ARGUMENTS. Use the API design from backend-architect: [include API endpoints and data models from step 1]"
   - Ensure UI matches the backend API contract

3. **Test Coverage**
   - Use Task tool with subagent_type="test-automator"
   - Prompt: "Write comprehensive tests for: $ARGUMENTS. Cover both backend API endpoints: [from step 1] and frontend components: [from step 2]"

4. **Production Deployment**
   - Use Task tool with subagent_type="deployment-engineer"
   - Prompt: "Prepare production deployment for: $ARGUMENTS..."

Analysis: This is explicit prompt-chaining by injection: [include API endpoints and data models from step 1] is an instruction to Claude to literally insert prior output as part of the next agent's prompt. The model: claude-opus-4-1 frontmatter pins the orchestrator to Opus.


Excerpt 2: security-scan.md (tools/)

Technique: Decision matrix + tool selection guide embedded in prompt (structured knowledge encoding)

---
model: claude-sonnet-4-0
---

# Security Scan and Vulnerability Assessment

You are a security expert specializing in application security, vulnerability assessment, and secure coding practices...

### 1. Security Scanning Tool Selection

Choose appropriate security scanning tools based on your technology stack:

**Tool Selection Matrix**
```python
security_tools = {
    'python': {
        'sast': {
            'bandit': {
                'strengths': ['Built for Python', 'Fast', 'Good defaults', 'AST-based'],
                'command': 'bandit -r . -f json -o bandit-report.json',
            },
            'semgrep': {
                'strengths': ['Multi-language', 'Custom rules', 'Low false positives'],
                'command': 'semgrep --config=auto --json --output=semgrep-report.json',
            }
        },
        'dependency_scan': {
            'safety': {
                'command': 'safety check --json --output safety-report.json',
            }
        }
    }

**Analysis**: Python dict literal embedded as knowledge schema within a markdown prompt — avoids prose repetition by encoding tool options in a structured format Claude can navigate algorithmically. Model pinned to Sonnet (not Opus) — appropriate for a scan task vs architectural reasoning.
09

Uniqueness

wshobson/commands — Uniqueness & Positioning

Differs From Seeds

Most similar to openspec (command+skill pairs), but wshobson/commands ships only commands (no skill counterparts). The workflows use a unique prompt-chaining pattern absent from seeds: instead of spawning anonymous subagents, workflow markdown files name specific subagent_type values that must exist in the user's environment — a separation of orchestration (this repo) from domain expertise (wshobson/agents). This is closer to taskmaster-ai's prompt-chaining pattern (where one stage's output becomes the next stage's prompt input) but implemented in pure slash-command Markdown rather than an MCP server. No seed uses explicit Task-tool invocation instructions inside markdown command files.

Historical Significance

This repo represents the "slash-command-first" era of Claude Code tooling (pre-plugin-marketplace). The README's explicit "consider the Plugin Marketplace instead" note makes this a useful historical artifact for tracking the evolution from command-centric to plugin-centric organization.

Observable Failure Modes

  1. Missing subagents: Workflow commands reference subagent_type values (backend-architect, etc.) that may not exist in the user's environment, causing silent failures.
  2. Model pinning drift: Hardcoded model names (claude-opus-4-1, claude-sonnet-4-0) will break as models are deprecated.
  3. No validation: Unlike wshobson/agents, there are no structural validators or CI checks.

Explicit Antipatterns

None stated. Implicit from README's deprecation note: the flat command organization without plugin boundaries is the antipattern this repo superseded.

04

Workflow

wshobson/commands — Workflow

User Workflow

  1. Clone to ~/.claude or project .claude/
  2. Invoke command: /workflows:feature-development <description>
  3. Command markdown instructs Claude Code to spawn specialized agents via Task tool
  4. Agents execute sequentially; each receives output from previous as context
  5. Final output is integration of all agent contributions

Feature-Development Workflow (Example)

Traditional Mode

  1. Backend Architecture Design — Task tool → subagent_type="backend-architect" — outputs API design + schema
  2. Frontend Implementation — Task tool → subagent_type="frontend-developer" — receives API design from step 1
  3. Test Coverage — Task tool → subagent_type="test-automator" — covers backend API + frontend components
  4. Production Deployment — Task tool → subagent_type="deployment-engineer" — prepares CI/CD + containerization

TDD Mode

  1. Task tool → subagent_type="tdd-orchestrator" — red-green-refactor cycle

TDD Commands Workflow

Separate tool commands for each TDD phase:

  • /tools:tdd-red — write failing test
  • /tools:tdd-green — implement minimum to pass
  • /tools:tdd-refactor — refactor while keeping tests green

Approval Gates

None explicit. Workflows proceed automatically through agent sequence.

Phase-to-Artifact Map

Phase Artifact
backend-architect API design doc + schema
frontend-developer UI components
test-automator Test suite
deployment-engineer CI/CD pipeline + Dockerfile
06

Memory Context

wshobson/commands — Memory & Context

State Storage

No persistent state. Context is maintained within a single session through Claude Code's native context window.

Context-Save / Context-Restore Commands

Two dedicated tools provide manual context persistence:

  • /tools:context-save — saves current session context to file
  • /tools:context-restore — restores previously saved context

These are basic file-based persistence commands, not a formal memory system.

Cross-Session Handoff

None beyond the context-save/restore pair.

Prompt-Chaining Context

In workflow commands (e.g., feature-development), each agent in the chain is instructed to incorporate prior agents' outputs as inline context in subsequent Task tool invocations. This is session-scoped only.

07

Orchestration

wshobson/commands — Orchestration

Multi-Agent Support

Yes — 15 workflow commands explicitly orchestrate multiple named subagents via Task tool.

Orchestration Pattern

Sequential (primary): agents execute in defined order with output hand-off. Example: backend-architect → frontend-developer → test-automator → deployment-engineer.

Task-decomposition-tree (TDD cycle): tdd-orchestrator subdivides into red→green→refactor.

Subagent References

Workflow commands reference subagent types by name (e.g., subagent_type="backend-architect", subagent_type="tdd-orchestrator") but these subagents are NOT shipped in this repo — they are expected to exist in the user's Claude Code environment (from wshobson/agents plugin marketplace or equivalent).

Isolation Mechanism

Process-level (Claude Code native subagent context isolation).

Multi-Model

No explicit model routing in workflow commands. Tool commands pin specific models in frontmatter (e.g., model: claude-opus-4-1 for feature-development, model: claude-sonnet-4-0 for security-scan).

Execution Mode

One-shot per command invocation.

Consensus

None.

Prompt Chaining

Yes — explicit instruction to inject prior agent outputs as inline context in subsequent agent prompts.

08

Ui Cli Surface

wshobson/commands — UI / CLI Surface

Dedicated CLI

None. Installation is via git clone.

Local Dashboard

None.

IDE Integration

Claude Code only (native slash command invocation).

Observability

None built-in. No audit log, no drift detection.

Invocation Patterns

# With directory prefix (recommended)
/workflows:feature-development implement OAuth2 authentication
/tools:security-scan perform vulnerability assessment

# Without prefix (copy files to root first)
/feature-development implement payment system
/api-scaffold create user management endpoints with RBAC

Related frameworks

same archetype · same primary tool · same memory type

claude-mem (thedotmack) ★ 78k

Background worker service captures every tool call as an observation, AI-compresses sessions, and auto-injects relevant past…

pi (badlogic/earendil) ★ 55k

A minimal, hackable, multi-provider terminal coding agent that adapts to your workflows via npm-installable TypeScript Extensions…

Agent Skills (Addy Osmani) ★ 46k

Encodes senior-engineer software development lifecycle as 23 auto-routed skills and 7 slash commands for any AI coding agent.

wshobson/agents Plugin Marketplace ★ 36k

Single Markdown source for 83 domain-specialized plugins that auto-generates idiomatic artifacts for five AI coding harnesses.

TabbyML/Tabby ★ 34k

Self-hosted AI coding assistant server (alternative to GitHub Copilot) with admin dashboard, RAG-based completions, and multi-IDE…

Compound Engineering ★ 17k

Make each unit of engineering work compound into easier future work via brainstorm→plan→execute→review→learn cycles.