Production-Grade Agent Skills for the Full Development Lifecycle

AI coding agents default to the shortest path. They skip specs, avoid writing tests, and push code that works but is not production-ready. Addy Osmani's agent-skills fixes this with 19 structured workflows that enforce the same discipline senior engineers bring to production code.

What Makes This Collection Different

Most skill repos give you individual tools: a commit writer, a code reviewer, a test generator. This one covers the entire software development lifecycle in six phases:

The skills draw from Google's engineering culture: Hyrum's Law in API design, the Beyonce Rule in testing, Chesterton's Fence in simplification, trunk-based development in git workflow.

Installation

# Claude Code marketplace
claude plugin marketplace add addyosmani/agent-skills
claude plugin install agent-skills@addy-agent-skills

# Or clone locally
git clone https://github.com/addyosmani/agent-skills.git

Skills activate automatically based on context. Building UI triggers frontend-ui-engineering. Designing an API triggers api-and-interface-design. You do not need to invoke them manually.

Real Workflow: Ship a Feature from Idea to Production

Step 1: Define the feature

/spec Build a user notification system that supports email, in-app, and push channels.

The spec-driven-development skill writes a PRD covering objectives, commands, structure, code style, testing strategy, and boundaries. The output is a document you review before any code is written.

Step 2: Break it into tasks

/plan Generate implementable tasks from this spec

The planning-and-task-breakdown skill decomposes the spec into small tasks with acceptance criteria and dependency ordering. Each task should be verifiable in isolation.

Step 3: Build incrementally

/build Implement the notification dispatch module first

The incremental-implementation skill enforces thin vertical slices: implement, test, verify, commit. It uses feature flags and safe defaults so every commit is deployable.

Step 4: Write tests first

/test Write tests for the notification dispatch module

The test-driven-development skill follows Red-Green-Refactor. It enforces the test pyramid (80% unit, 15% integration, 5% E2E) and the Beyonce Rule: if it can break in production, it needs a test.

Step 5: Review before merge

/review Review the notification feature changes

The code-review-and-quality skill runs a five-axis review: correctness, readability, security, performance, and maintainability. It sizes changes (~100 lines per review) and labels findings by severity.

Step 6: Ship safely

/ship Deploy the notification feature

The shipping-and-launch skill runs a pre-launch checklist, sets up staged rollouts via feature flags, and configures monitoring with rollback procedures.

Slash Commands

The collection includes 7 commands that chain skills together:

When Not to Use This

Skip this if you want lightweight, single-purpose skills. This collection enforces process. If you just need a commit message writer or a quick code review, the official Anthropic skills are simpler. This collection is for teams that want structured engineering discipline baked into their agent workflow.

Tips

• Start with /spec on your next feature. The discipline of writing a spec first changes how you think about the problem.
• Use /code-simplify on existing code. It applies Chesterton's Fence: understand why code exists before removing it.
• The security-and-hardening skill is worth running on any code that handles user input, authentication, or external data.
• Keep change sizes under 100 lines. The skills enforce this, but it works best when you follow along.

Looking for more engineering skills? Check the SkillMap leaderboard for the latest curated collection.