Your dev workflow, automated end to end.
An opinionated development workflow for Claude Code — customizable to your opinions.
Install •
Quick Start •
Commands •
Config •
Extending
Why Hero Skills?
Most dev work follows the same loop: grab a ticket, plan, implement, test, review, commit, push, monitor. But every team does it slightly differently — different PM tools, different CI, different deploy targets.
Hero Skills gives you slash commands for the entire dev lifecycle that adapt to your stack. Configure once with HERO.md, then every skill knows your conventions, your tools, and your preferences.
- Plan from tickets — Fetch from Linear/Jira/GitHub Issues, create branches, draft implementation plans
- Implement with guardrails — Execute plans step by step with per-step verification
- Test anything — Auto-detect project type (API, frontend, CLI, MCP) and run the right smoke tests
- Ship with confidence — Self-review, pre-commit checks, conventional commits, PR creation
- Stay informed — CI/CD status, cluster health, security scans
Install
git clone https://github.com/ai-hero/hero-skills.git ~/.claude/plugins/hero-skills
Skills are immediately available in any Claude Code session. No restart needed.
Quick Start
# 1. Configure your project (run once per repo)
/hero-init
# 2. Plan work from a ticket
/hero-plan PROJ-123
# 3. Implement the approved plan
/hero-implement
# 4. Test, review, commit, push
/hero-test
/hero-commit
/hero-push
That's it. Each command reads your HERO.md config and adapts to your stack automatically.
Commands
Setup
| Command | What it does |
|---|
/hero-init | Investigate your repo, auto-detect stack, create HERO.md config |
/hero-setup | Set up a developer's local environment (tools, auth, dependencies) |
/hero-new | Scaffold a new project (Python, full-stack, Node.js) in any repo structure |
Development Cycle
| Command | What it does |
|---|
/hero-plan | Fetch a ticket, create a branch, draft an implementation plan |
/hero-implement | Execute an approved plan step by step with verification |
/hero-test | Auto-detect project type and run smoke tests |
/hero-commit | Self-review, pre-commit checks, grouped conventional commits |
/hero-push | Push, create PR with generated description, or merge |
Code Review
| Command | What it does |
|---|
/hero-review-pr | Review a PR and leave inline comments (quality, bugs, security, style) |
/hero-respond-to-pr | Fix PR review comments, resolve threads, optionally loop with external review agent |
Operations
| Command | What it does |
|---|
/hero-cicd | Check CI/CD pipeline status, build logs, image publish status |
/hero-health | Kubernetes cluster health (nodes, pods, deployments, ArgoCD) |
/hero-secure | Scan dependencies (Dependabot) and containers (Docker Scout) for CVEs |
Architecture & Meta
| Command | What it does |
|---|
/hero-architect | Generate architecture specs with Mermaid diagrams |
/hero-new-skill | Create new Claude Code skills, rules, or hooks |
/hero-meta | Audit the hero-skills plugin itself for quality and consistency |
HERO.md
Every skill reads HERO.md from your repo root. It declares your stack so skills don't have to guess. HERO.md is committed to the repo — it's team-shared, so every developer and every hero skill works from the same config.
To keep it in sync automatically, wire /hero-init --update into your pre-commit hooks. A fast bash gate script checks staged files first — most commits skip Claude entirely and finish in milliseconds. Only when you change dependencies, CI config, or project structure does it invoke Claude to sync HERO.md.
Here's what a minimal config looks like:
# HERO Configuration
## Project Management
- Tool: Linear
- Project: PROJ
## CI/CD
- Platform: GitHub Actions
## Code Quality
- Pre-commit: true
- Formatter: ruff format
- Linter: ruff check
## Projects
### api
- Language: Python
- Framework: FastAPI
- Test command: pytest
- Dev command: uvicorn main:app --reload
ai-hero