Skip to content

efimsky/ai-code-ninja

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

ai-code-ninja

License: Apache 2.0 Claude Code

Custom slash commands and skills for Claude Code - Anthropic's AI coding assistant CLI. Automate GitHub issue workflows, codebase exploration, and session handoffs.

A small collection of battle-tested Claude Code commands and skills for structured development workflows.

Why use it

  • Consistent, phase-based workflow for issues and PRs
  • Safer changes with explicit planning and approval gates
  • Run multiple Claude Code sessions concurrently without stepping on each other
  • Less busywork: worktrees, issue comments, and PR creation are automated

Install

  1. Global (all projects) — commands and skills
git clone https://github.com/efimsky/ai-code-ninja.git
mkdir -p ~/.claude/commands ~/.claude/skills
cp ai-code-ninja/commands/*.md ~/.claude/commands/
cp -r ai-code-ninja/skills/* ~/.claude/skills/
  1. Project-only
mkdir -p .claude/commands .claude/skills
cp path/to/ai-code-ninja/commands/*.md .claude/commands/
cp -r path/to/ai-code-ninja/skills/* .claude/skills/
  1. Single command (without skills)
curl -o ~/.claude/commands/gh_issue.md \
  https://raw.githubusercontent.com/efimsky/ai-code-ninja/main/commands/gh_issue.md

💡 /gh_issue references the verification-before-completion skill — install the skills directory alongside the commands for the full quality-gate behavior. Without the skill, the command falls back to an inline summary of the rule.

Quick start

git clone https://github.com/efimsky/ai-code-ninja.git
mkdir -p ~/.claude/commands
cp ai-code-ninja/commands/*.md ~/.claude/commands/

# In Claude Code
/gh_issue 42

Commands

/gh_issue <ISSUE_URL_OR_NUMBER>

Complete GitHub issue workflow with structured phases, quality gates, and progress tracking.

Workflow: validation -> research -> planning -> approval -> implementation -> QA -> testing -> PR -> cleanup

Key features:

  • Creates isolated git worktrees for each issue
  • Posts implementation plans as GitHub comments
  • Waits for explicit approval before implementing
  • Follows conventional commit format
  • Handles merge conflicts and error recovery

Usage:

/gh_issue 42
/gh_issue https://github.com/owner/repo/issues/42

/gh_issue_review <ISSUE_URL_OR_NUMBER>

Review and refine GitHub issues with multiple SME perspectives before implementation.

Workflow: validation -> context gathering -> SME analysis -> user interview -> restructure -> approval -> update

Key features:

  • Applies PM, BA, Tech Lead, and Test Lead perspectives
  • Explores codebase for context-aware feedback
  • Searches for similar/duplicate issues
  • Comprehensive user interview to fill gaps
  • Adapts issue format based on type (bug/feature/enhancement)
  • Requires explicit approval before updating the issue

Usage:

/gh_issue_review 42
/gh_issue_review https://github.com/owner/repo/issues/42

/explore <TOPIC_OR_QUESTION>

Guided codebase exploration for understanding code before implementation. Strictly read-only.

Key features:

  • Uses Explore agents for thorough search
  • Asks clarifying questions to understand your goal
  • Produces structured output with file:line references
  • Output options: console, .claude/explorations/, or GitHub issue
  • Offers follow-up questions for deeper exploration

Usage:

/explore authentication
/explore "how do API endpoints handle errors"
/explore state management

/handoff

Capture current session state for seamless continuation in a new conversation.

Key features:

  • Auto-detects /gh_issue workflows via branch pattern
  • Captures git branch, uncommitted changes, and todo state
  • Generates structured markdown for pasting into new session
  • Includes continuation prompt for next session

Usage:

/handoff

Skills

verification-before-completion

Always-on quality gate that fires whenever Claude is about to claim work is "done", "fixed", "tests pass", "ready", etc. Enforces a four-step proof: name the proving command, run it fresh, read full output and exit code, match the output to the claim.

Why it's a skill (not just part of /gh_issue):

  • Auto-triggers across all your work, not just GitHub issue workflows
  • Installed once, applies in ad-hoc bug-fix conversations, refactors, and one-off scripts
  • Includes a required-evidence-by-claim-type table and an extensive rationalization-rejection table

Install (global):

mkdir -p ~/.claude/skills
cp -r ai-code-ninja/skills/* ~/.claude/skills/

Inspired by the Superpowers pattern of pulling cross-cutting discipline rules out of workflows and into composable skills.

Requirements (for /gh_issue)

Required:

  • GitHub CLI (gh) - issue fetching, PR creation, and all GitHub interactions (install guide)
  • code-review plugin - Phase 5 code review (claude plugins:install code-review)
  • code-simplifier plugin - Phase 5 code simplification (claude plugins:install code-simplifier)

Recommended:

  • verification-before-completion skill (bundled in this repo's skills/ directory) - quality gate that catches false-completion claims

Optional:

  • frontend-design plugin - UI/UX planning for frontend issues (claude plugins:install frontend-design)
  • Chrome for Claude - UI screenshots/GIFs (extension)

Also requires:

  • A Git repository with a GitHub remote

Quick setup:

claude plugins:install code-review code-simplifier frontend-design

YOLO Mode (optional)

Skip permission prompts for git/gh commands by copying the included settings to your project:

cp ai-code-ninja/.claude/settings.json .claude/

Or add to your global ~/.claude/settings.json:

{
  "permissions": {
    "allow": [
      "Bash(git worktree:*)",
      "Bash(git fetch:*)",
      "Bash(git rebase:*)",
      "Bash(git push:*)",
      "Bash(git branch -D:*)",
      "Bash(git revert:*)",
      "Bash(gh issue close:*)",
      "Bash(gh issue comment:*)",
      "Bash(gh issue create:*)",
      "Bash(gh issue edit:*)",
      "Bash(gh issue view:*)",
      "Bash(gh pr create:*)"
    ]
  }
}

Customization

Commands are markdown files with YAML frontmatter. You can customize:

  • model - Force opus/sonnet/haiku for the command
  • allowed-tools - Restrict which tools the command can use
  • Phases and workflows to match your team's process

See CONTRIBUTING.md for the full command format.

Contributing

Contributions welcome! See CONTRIBUTING.md for guidelines on:

  • Command file structure
  • Quality guidelines
  • Submitting new commands

License

MIT

About

Custom Claude Code commands for GitHub issue workflows, PR automation, and developer productivity

Resources

License

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors