SlashAgents

Version: 1.4.0 | Updated: 2025-01-11 | Author: Ricardo Carvalho

A curated collection of 53 battle-tested Claude Code extensions designed to help developers stay in flow. This toolkit includes 16 slash commands (workflow automation like /pr and /ci-orchestrate), 35 agents (specialized agents for testing, code quality, BMAD workflows, and automation), and 2 skills (reusable PR and refactoring operations).

Whether you're managing pull requests, orchestrating CI pipelines, or following structured BMAD development cycles, these tools preserve your creative momentum by automating repetitive tasks and providing intelligent assistance exactly when you need it.

Contents

Type	Count	Description
Commands	16	Slash commands for workflows (/pr, /ci-orchestrate, etc.)
Agents	35	Specialized agents for testing, quality, BMAD, and automation
Skills	2	Reusable skill definitions (PR workflows, safe refactoring)

Note: Plugin installation includes 16 command aliases for convenience (e.g., /pr instead of /cc-agents-commands:pr).

Prerequisites

Before installing CC_Agents_Commands, ensure you have:

• Required:

  • Claude Code CLI installed and working

• Optional (enhances specific tools):

  • MCP servers - Some commands require MCP servers (e.g., github MCP for /pr and /ci-orchestrate). See plugin/MCP_SETUP.md for detailed MCP server configuration.
  • BMAD framework - Required for epic development workflows (/epic-dev, /epic-dev-full, /epic-dev-init). Install from BMAD repository.

Installation

Choose your installation method based on your needs:

Method	Command	Scope	Pros	Cons	Best For
Plugin Install (Recommended)	claude --plugin-dir ./plugin	Project-local	Single command, includes MCP configs, auto-loads aliases, includes utility scripts	Requires staying in project directory	Most users, quick setup, full feature access
File Install (Alternative)	cp -r commands/ ~/.claude/commands/ etc.	Global or project	Full control, works anywhere, no plugin dependencies	Manual copy steps, no MCP auto-config, no aliases	Power users, custom setups, selective tool installation

Note: Plugin Install automatically configures MCP servers and command aliases. File Install requires manual MCP setup.

Recommended: Plugin Install

The easiest way to get started with all features enabled:

1. Clone the repository

   git clone https://github.com/Autopsias/slashagents.git
   cd slashagents

2. Install as plugin

   claude --plugin-dir ./plugin

   This single command loads all 53 extensions (16 commands, 35 agents, 2 skills) plus:

   • 16 command aliases for convenience (e.g., /pr instead of /cc-agents-commands:pr)
   • MCP server configurations (.mcp.json)
   • Event hooks (hooks.json)
   • Utility scripts

3. Verify installation

   In your Claude Code session, type /help and you should see the installed commands.

Tip: For MCP server configuration and advanced features, see plugin/MCP_SETUP.md.

Alternative: File Install (Power Users)

For users who prefer manual installation with full control:

⚠️ Warning: If you have existing commands in ~/.claude/, back them up first. New files with the same name will overwrite existing ones.

1. Clone the repository

   git clone https://github.com/Autopsias/slashagents.git
   cd slashagents

2. Backup existing commands (if any exist)

   cp -r ~/.claude ~/.claude.backup

3. Copy tools to your target location

   Choose between global installation (tools available in all projects) or project installation (project-specific, version-controlled).

   Global installation (~/.claude/) - recommended for personal productivity tools:

   cp -r commands/ ~/.claude/commands/
   cp -r agents/ ~/.claude/agents/
   cp -r skills/ ~/.claude/skills/

   Project installation (.claude/) - recommended for team projects:

   cp -r commands/ .claude/commands/
   cp -r agents/ .claude/agents/
   cp -r skills/ .claude/skills/

4. Start a new Claude Code session

   Commands load automatically when you start a new session (no restart required).

5. Verify installation

   In your Claude Code session, type /help and you should see your installed commands listed.

Note: File Install does not include MCP configurations or command aliases. You'll need to configure MCP servers manually and use full command names (e.g., /commit-orchestrate instead of /commit).

Quick Start

Try these commands to experience immediate value:

Resume Your Workflow

Generates a continuation prompt to pick up where you left off in a new session.

/nextsession

Check PR Status

Shows your open pull requests and their current state. Note: Enhanced with github MCP server for full functionality.

/pr status

Orchestrate Quality Checks

Runs quality checks, stages changes, and creates a well-formatted commit.

/commit-orchestrate

The 'Aha' Moment: Auto-Fix CI Failures

Analyzes CI failures, spawns parallel agents, and fixes issues automatically. Note: Enhanced with github MCP server for full functionality.

/ci-orchestrate

Ralph Loop (Unattended Execution)

Several commands support Ralph Loop mode for unattended/overnight execution. This pattern spawns fresh Claude instances per iteration, preventing context exhaustion on long-running tasks.

Usage

Add --loop N to any supported command:

# Run epic development overnight (max 10 iterations)
/epic-dev 2 --loop 10

# Run code quality fixes with 30s delay between iterations
/code-quality --fix --loop 5 --loop-delay 30

# CI fixes in loop mode
/ci-orchestrate --loop 10

Supported Commands

Command	Completion Signal	Granularity
/epic-dev	Epic complete	Phase-level (CREATE → DEVELOP → REVIEW)
/epic-dev-full	Epic complete	Phase-level (8 phases: CREATE → VALIDATION → ATDD → DEV → REVIEW → AUTOMATE → TEST_REVIEW → TRACE)
/code-quality	All violations fixed	Rule-level (complexity → file-length → duplication)
/test-orchestrate	All tests passing	Type-level (unit → integration → e2e)
/ci-orchestrate	All CI checks passing	Category-level (linting → types → tests)

The --loop N Parameter

N is the maximum iterations, not the exact count:

• Loop exits early on completion signals (e.g., "All tests passing", "Epic complete")
• Loop exits on blocking signals requiring human intervention (e.g., "HALT", "Manual intervention required")
• If neither signal detected after N iterations, loop ends with partial completion

Think of N as a safety limit for unattended/overnight runs.

Signal Type	Example	Behavior
Completion	All tests passing, Epic complete	✅ Exit with success
Blocking	HALT, Manual intervention required	⚠️ Exit, needs human
Max reached	Iteration N completes	⚠️ Exit, may have remaining work

How It Works

1. Spawns a fresh Claude instance per iteration (full 200K context)
2. Executes one phase/category per iteration (not entire workflow)
3. Detects completion signals to exit early on success
4. Detects blocking signals to halt for human intervention
5. Configurable max iterations (--loop N) and delay (--loop-delay S)

Phase-Level Granularity (v1.5.0+)

Ralph loops now operate at phase level instead of story/task level for improved token efficiency and fresh perspective per phase:

Aspect	Before (Story-Level)	After (Phase-Level)
Token Cost	150-200K per iteration	20-50K per iteration
Context	Accumulated (tunnel vision)	Fresh per phase
Checkpoints	1 per story	1 per phase
User Intervention	Between stories only	Between phases

Example flow with /epic-dev 2 --loop 10:

Iteration 1: [PHASE: CREATE] Creating story 2-1-auth
Iteration 2: [PHASE: DEVELOP] Implementing story 2-1-auth (with Gate 2.5 verification)
Iteration 3: [PHASE: REVIEW] Reviewing story 2-1-auth (with Gate 3.5 verification)
Iteration 4: [PHASE: CREATE] Creating story 2-2-profile
...

Each iteration gets fresh context, preventing accumulated confusion and enabling better recovery from failures.

Attribution: Pattern inspired by snarktank/ralph

CI/CD

This project uses self-hosted macOS runners for continuous integration. See docs/SELF_HOSTED_RUNNERS.md for runner setup and maintenance instructions.

Pipeline: .github/workflows/docs-ci.yml

• Markdown linting (markdownlint-cli2)
• Link validation (markdown-link-check)
• Shell script linting (ShellCheck)
• File structure validation
• Content validation (frontmatter, hardcoded paths)

Maintenance: Weekly automated cleanup via runner-maintenance.yml

Commands Reference

Commands are organized by workflow moment to help you quickly find the right tool for your task.

Shared Libraries

Commands in commands/lib/ provide reusable patterns used internally by orchestrators:

Library	Purpose
ralph-loop.md	Fresh-context loop pattern for unattended execution
status-updater.md	Reliable status file update patterns for orchestrators

Starting Work

Command	What it does	Prerequisites
/nextsession	Generates continuation prompt for next session	—
/epic-dev-init	Verifies BMAD project setup for epic development	BMAD framework

Building

Command	What it does	Prerequisites
/epic-dev	Automates BMAD development cycle for epic stories (supports --loop)	BMAD framework
/epic-dev-full	Executes full TDD/ATDD-driven BMAD development (supports --loop)	BMAD framework
/epic-dev-epic-end-tests	Validates epic completion with NFR assessment	BMAD framework
/parallel	Smart parallelization with file conflict detection and specialist routing	—

Quality Gates

Command	What it does	Prerequisites
/ci-orchestrate	Orchestrates CI failure analysis and fixes (supports --loop)	github MCP
/test-orchestrate	Orchestrates test failure analysis and fixes (supports --loop)	test files and results
/code-quality	Analyzes and fixes code quality issues (supports --loop)	code files in project
/coverage	Orchestrates test coverage improvement	test coverage tools
/create-test-plan	Creates comprehensive test plans	project documentation
/test-epic-full	Tests epic-dev-full command workflow	BMAD framework
/user-testing	Facilitates user testing sessions (uses interactive-guide agent)	user testing setup
/usertestgates	Finds and runs next test gate (uses evidence-collector agent)	test gates in project

Shipping

Command	What it does	Prerequisites
/pr	Manages pull request workflows	github MCP
/commit-orchestrate	Orchestrates git commit with quality checks	—

Agents Reference

Agents are organized by domain to help you quickly find the right specialist for your task.

Test Fixers

Agent	What it does	Prerequisites
unit-test-fixer	Fixes Python test failures for pytest and unittest	test files in project
api-test-fixer	Fixes API endpoint test failures	API test files in project
database-test-fixer	Fixes database mock and integration test failures	database test files in project
e2e-test-fixer	Fixes Playwright E2E test failures	E2E test files in project

Code Quality

Agent	What it does	Prerequisites
linting-fixer	Fixes Python linting and formatting issues	linting config in project
type-error-fixer	Fixes Python type errors and annotations	Python/TypeScript project
import-error-fixer	Fixes Python import and dependency errors	code files in project
security-scanner	Scans code for security vulnerabilities	code files in project
code-quality-analyzer	Analyzes code quality metrics and patterns	code files in project
requirements-analyzer	Analyzes and validates project requirements	project documentation

Workflow Support

Agent	What it does	Prerequisites
pr-workflow-manager	Manages pull request workflows via GitHub	github MCP
parallel-orchestrator	Spawns parallel agents with conflict detection	—
digdeep	Performs Five Whys root cause analysis	perplexity-ask MCP

Testing & Strategy

Tip: Use /user-testing and /usertestgates commands to orchestrate these agents for user acceptance testing.

Agent	What it does	Prerequisites
test-strategy-analyst	Analyzes test failures with Five Whys methodology	perplexity-ask MCP, exa MCP
test-documentation-generator	Generates test failure runbooks and documentation	test files in project
ui-test-discovery	Discovers UI components for test generation	UI code in project
validation-planner	Plans validation strategies for features	project files
scenario-designer	Transforms requirements into test scenarios	project files
evidence-collector	Validates and collects test evidence (used by /usertestgates)	project files
interactive-guide	Guides human testers through validation (used by /user-testing)	project files

BMAD Workflow

Agent	What it does	Prerequisites
epic-story-creator	Creates user stories from epics	BMAD framework
epic-story-validator	Validates stories and quality gates	BMAD framework
epic-test-generator	Generates ATDD tests for stories	BMAD framework
epic-atdd-writer	Generates failing acceptance tests (TDD RED phase)	BMAD framework
epic-implementer	Implements stories (TDD GREEN phase)	BMAD framework
epic-test-expander	Expands test coverage after implementation	BMAD framework
epic-test-reviewer	Reviews test quality against best practices	BMAD framework
epic-code-reviewer	Adversarial code review (finds 3-10 issues)	BMAD framework

CI/DevOps

Agent	What it does	Prerequisites
ci-strategy-analyst	Analyzes CI/CD pipeline issues strategically	perplexity-ask MCP, exa MCP
ci-infrastructure-builder	Builds CI/CD infrastructure and workflows	github MCP
ci-documentation-generator	Generates CI/CD documentation	CI infrastructure

Browser Automation

| Agent | What it does | Prerequisites |

| ------- | -------------- | --------------- |

| browser-executor | Executes browser automation with Chrome DevTools | chrome-devtools MCP or playwright MCP |

| chrome-browser-executor | Chrome-specific browser automation executor | chrome-devtools MCP |

| playwright-browser-executor | Playwright-specific browser automation executor | playwright MCP |

Skills Reference

Skills leverage agents to provide natural language interfaces for complex multi-step workflows.

Skill	What it does	Prerequisites
pr-workflow	Manages PR workflows - create, status, merge, sync	github MCP (via pr-workflow-manager)
safe-refactor	Test-safe file refactoring with facade pattern	code files in project

Troubleshooting

Prerequisites notation: The Prerequisites column uses — for standalone tools, server-name MCP for tools requiring MCP servers, and "BMAD framework" or descriptive text for specialized requirements.

MCP server not working? Configure the required MCP server in your Claude settings (e.g., github or perplexity-ask). Check the Prerequisites column in reference tables above to see which tools require MCP servers. Then restart Claude Code to apply the changes.

BMAD commands not found? Install the BMAD framework from BMAD repository before using /epic-dev, /epic-dev-full, or /epic-dev-init commands. See BMAD documentation for installation instructions.

Command or agent not recognized? Verify files are in the correct location (~/.claude/ for global or .claude/ for project). Then start a new Claude Code session (commands load automatically). Ensure you are in an active Claude Code session, not a regular terminal.

Requirements

• Claude Code CLI installed
• Some extensions require specific MCP servers (noted in individual files)
• BMAD extensions require BMAD framework installed

License

MIT