Production-grade skills
for AI coding agents.

Encode the workflows, quality gates, and judgment senior engineers bring to production - packaged so your agent follows them consistently across every phase, from first idea to live deploy.

npx skills add addyosmani/agent-skills

Get started Browse the 24 skills

Works with 70+ agents

Claude CodeCodexCursorCopilotWindsurfClineAntigravityOpenCodeKiro

01 The lifecycle

One command per phase. Skills load automatically.

Eight slash commands map to the way software actually gets built. Each one activates the right skills for the moment - designing an API pulls in contract-first design, building UI pulls in accessibility, and so on.

Define
Idea Refine
/spec
Plan
Spec PRD
/plan
Build
Code Impl
/build
Verify
Test Debug
/test
Review
QA Gate
/review
Ship
Go Live
/ship

/spec Define what to build Spec before code /plan Plan how to build it Small, atomic tasks /build Build incrementally One slice at a time /test Prove it works Tests are proof /review Review before merge Improve code health /webperf Audit web performance Measure before you optimize /code-simplify Simplify the code Clarity over cleverness /ship Ship to production Faster is safer

02 The catalog

24 skills spanning the whole SDLC

Not prompts - structured workflows with steps, checkpoints, and exit criteria. Reach for one directly, or let the meta-skill route you.

View all skills

Define

interview-me

One-question-at-a-time interview that extracts what you actually want.

The ask is underspecified, or you invoke "interview me" / "grill me".

Define /spec

spec-driven-development

Write a PRD covering objectives, structure, style, testing and boundaries - before code.

Starting a new project, feature, or significant change.

Build /build

incremental-implementation

Thin vertical slices - implement, test, verify, commit. Rollback-friendly changes.

Any change touching more than one file.

Build /test

test-driven-development

Red-Green-Refactor, the test pyramid, DAMP over DRY, the Beyoncé Rule.

Implementing logic, fixing bugs, or changing behavior.

Build

doubt-driven-development

Adversarial fresh-context review of every non-trivial decision, in-flight.

Stakes are high, code is unfamiliar, or verifying now beats debugging later.

Build

frontend-ui-engineering

Component architecture, design systems, state, responsive, WCAG 2.1 AA accessibility.

Building or modifying user-facing interfaces.

Review /review

code-review-and-quality

Five-axis review, ~100-line change sizing, severity labels, review-speed norms.

Before merging any change.

Review

security-and-hardening

OWASP Top 10 prevention, auth patterns, secrets, dependency auditing.

Handling user input, auth, data storage, or external integrations.

Review /webperf

performance-optimization

Measure-first - Core Web Vitals targets, profiling, bundle analysis, N+1 fixes.

Performance requirements exist or you suspect regressions.

Ship /ship

shipping-and-launch

Pre-launch checklists, feature-flag lifecycle, staged rollouts, rollback procedures.

Preparing to deploy to production.

03 Why it works

AI agents default to the shortest path.
These skills close the shortcut.

Left alone, agents skip specs, tests, and reviews - the practices that make software reliable. Four mechanisms hold the line.

Anti-rationalization

Agents stop talking their way out of the work

Every skill ships a table of the exact excuses an agent makes to skip a step - “I’ll add tests later,” “this is too simple to spec” - each one rebutted inline. The shortcut is closed before it’s taken.

Verification gates

Evidence, not “seems right”

Each skill ends with exit criteria: tests passing, build output, real runtime data. Nothing is “done” on vibes. Verification is the whole point, not a nicety bolted on at the end.

Parallel personas

A review panel on every ship

/ship fans out to a code-reviewer, security-auditor, test-engineer and web-performance-auditor in parallel, then merges their findings into a single go / no-go decision.

In-repo evals

The skills are measured, not asserted

A three-tier eval framework checks structure, routing (no two skills collide on the words users say), and real execution traces - some of it in CI. The catalog is held to its own standard.

04 Specialist review

Ship past a panel, not a rubber stamp

/ship runs a pre-launch checklist and fans out to four specialist personas in parallel. Each brings a different perspective; their findings merge into one honest go / no-go.

See the /ship flow

code-reviewer

Senior Staff Engineer

Five-axis code review with a "would a staff engineer approve this?" standard.

test-engineer

QA Specialist

Test strategy, coverage analysis, and the Prove-It pattern.

security-auditor

Security Engineer

Vulnerability detection, threat modeling, and OWASP assessment.

web-performance-auditor

Web Performance Engineer

Core Web Vitals audit with Quick/Deep modes and a metric-honesty rule. Run via /webperf.

Give your agent a senior engineer’s instincts.

One command installs all 24 skills into 70+ agents.