🎯Skills21
Production-grade engineering skills encoding senior-level workflows, quality gates, and best practices for every phase of the software development lifecycle.
Production-grade engineering skills for AI coding agents that enforce structured workflows, quality gates, and best practices across the full software development lifecycle including spec, planning, build, testing, review, and shipping.
A collection of production-grade engineering skills for AI coding agents that encode structured workflows, quality gates, and best practices across every phase of the software development lifecycle from spec writing to deployment.
Plan-mode-first skill that decomposes a spec into small, verifiable tasks with explicit acceptance criteria — covering task sizing, dependency ordering, checkpoint design, and writing tasks small enough for an agent to execute reliably. Part of Addy Osmani's Agent Skills collection encoding senior-engineer workflows across Define / Plan / Build / Verify / Review / Ship phases.
A production-grade agent skill from addyosmani/agent-skills that activates the /code-simplify command to guide AI agents toward clarity over cleverness when refactoring code.
Quality gates before merge.
Production-grade engineering skills for AI coding agents that encode senior-engineer workflows, quality gates, and best practices across the full software development lifecycle from spec to shipping.
Safe, documented, reversible releases.
Refined ideas and structured requirements before code.
Small slices, right context, clean interfaces.
Prove it works with tests and real browser data.
Production-grade engineering skills for AI coding agents.
Small slices, right context, clean interfaces.
Prove it works with tests and real browser data.
Small slices, right context, clean interfaces.
An agent skill for browser testing with DevTools, part of a production-grade engineering skills collection that encodes senior engineer workflows and quality gates for every phase of the software development lifecycle.
Safe, documented, reversible releases.
Sets up and modifies CI/CD pipelines around a strict shift-left quality gate (lint → typecheck → unit tests → build → integration → optional E2E → security audit → bundle size), with ready-to-paste GitHub Actions workflows for the basic pipeline, Postgres-backed integration tests, Playwright E2E, preview deploys, feature-flag rollouts, and rollback workflows. Includes guidance on feeding CI failure output back into the agent loop, environment/secret separation, Dependabot, and the "Build Cop" role.
Production-launch playbook covering a full pre-launch checklist (code quality, security, performance, accessibility, infra, docs), feature-flag-gated staged rollout (deploy off → team → 5% canary → 25/50/100), explicit rollout decision thresholds (error rate, p95 latency, client errors, business metrics), monitoring/observability setup, post-launch verification, and a written rollback plan. Use when preparing any production deploy, opening a beta, or migrating data/infrastructure.
Treats code as a liability and walks through deprecation/migration end-to-end: when to maintain vs sunset (decision questions), advisory vs compulsory deprecation, the migration process (build replacement → announce/document → migrate consumers incrementally → remove old system), and patterns like Strangler, Adapter, and feature-flag migration. Calls out Hyrum's Law, the Churn Rule, zombie code, and common rationalizations to push back on.
Ground every framework-specific implementation decision in official documentation rather than training data. Workflow: Detect stack/versions from `package.json`/`pyproject.toml`/`go.mod`/etc., fetch the exact relevant docs page (e.g. `react.dev/reference/react/useActionState`, not the homepage), implement matching those patterns, and cite full URLs (prefer deep links with anchors) in code comments. Treats Stack Overflow, blog posts, and training data as non-authoritative and requires explicit `UNVERIFIED:` flags when no doc exists.