Skills for Real Engineers Matt Pocock AIHero Curated Resource and Installation Guide CookApps CookApps Marketplace

Skills for Real Engineers is the definitive agent skills library created by Matt Pocock (creator of Total TypeScript and founder of AI Hero). Built from decades of production software engineering experience, this suite contains 39 composable, model-agnostic skills that transform coding agents from superficial "vibe coding" generators into disciplined, senior-level software engineers.

Most AI-assisted development tools fail because they take away developer control, accelerate software entropy, or hallucinate mismatched implementations. Matt Pocock's skills solve these failure modes by introducing active domain modeling, relentless alignment interviews (grilling), test-driven development (TDD) with strict seams, and continuous codebase architecture deepening.

Overview and Core Capabilities

Modern coding agents like Claude Code, OpenAI Codex, Cursor, and Google Antigravity have immense token throughput, but they frequently encounter four critical failure modes in real production codebases:

The Agent Did Not Do What I Want: Misalignment between human intent and AI assumptions. Solved through rigorous grilling sessions that interrogate every design branch before writing code.

The Agent Is Way Too Verbose: Agents struggle with project-specific jargon, generating excessive explanation tokens. Solved by building a shared ubiquitous language in GLOSSARY.md and recording Architecture Decision Records (ADRs).

The Code Does Not Work: Lack of feedback loops causes agents to produce broken implementations. Solved through strict red-green-refactor TDD cycles and disciplined bug diagnosis loops.

The Codebase Becomes a Ball of Mud: Rapid AI generation accelerates software entropy. Solved through architectural surveys that actively detect shallow modules and refactor them into deep, testable interfaces.

Every skill is designed to be lightweight, modular, and compatible across all major agent environments including Claude Code, Codex, GitHub Copilot, Gemini CLI, Cursor, and Antigravity.

Comprehensive Breakdown of the 39 Specialized Skills

The repository organizes its 39 skills into five cohesive engineering and productivity modules:

1. Engineering User-Invoked Skills

Orchestration skills directly triggered by developer slash commands to guide the software lifecycle:

ask-matt: Intelligent routing assistant that assesses your current development context and directs you to the optimal skill or workflow.

grill-with-docs: Relentless alignment interview that sharpens feature specifications while actively building your project's domain model, updating GLOSSARY.md, and authoring ADRs inline.

triage: Moves issues and pull requests through a formal state machine with automated label management.

improve-codebase-architecture: Deepening survey that scans recent commit histories, detects shallow abstractions, renders an interactive HTML report, and guides refactoring.

setup-matt-pocock-skills: Repository initialization runbook that configures issue tracker bindings, triage label taxonomy, and domain documentation directories.

to-spec: Synthesizes the active conversation into a structured technical specification and publishes it directly to your issue tracker without unnecessary interrogation.

to-tickets: Breaks monolithic plans or specifications into tracer-bullet tickets with explicit dependency blocking edges.

implement: Executes ticketed work by driving test-driven development at verified seams and concluding with automated two-axis code reviews.

implement-spec: Concurrently implements an entire specification branch using parallel implementer subagents across the ready task graph.

wayfinder: Strategic mapping skill that breaks massive cross-session initiatives into decision tickets, resolving dependencies sequentially.

retro: Post-session retrospective assistant that audits navigation, automated checks, and coding standards, presenting prioritized environment improvements.

2. Engineering Model-Invoked Skills

Autonomous engineering practices reached for by the agent or invoked manually when disciplined execution is required:

prototype: Builds rapid throwaway prototypes (single shareable HTML or radical UI alternatives) to answer critical design and feasibility questions.

diagnosing-bugs: Systematic six-stage debugging protocol: build red reproduction, minimise, hypothesise, instrument, fix, and regression test.

research: Autonomous research agent that queries trusted primary sources and compiles cited Markdown documentation.

tdd: Test-driven development engine enforcing strict red-green-refactor loops at pre-agreed public seams.

domain-modeling: Active domain refinement discipline that challenges vocabulary against GLOSSARY.md, invents edge-case scenarios, and codifies ADRs.

codebase-design: Architectural vocabulary and principles for designing deep modules with high functionality behind minimal public interfaces.

code-review: Parallel two-axis code review auditing both coding standards (Fowler code smells) and specification fidelity.

pr: Formats pull request descriptions with concise visual summaries, before-and-after evidence, and merge risk blast radius evaluations.

wizard: Generates interactive bash execution wizards for complex manual developer tasks including cloud provisioning and credential setup.

3. Productivity and Communication Suite

Essential alignment and collaboration tools applicable to both engineering and product workflows:

grill-me: Relentless interview primitive that questions users on plans, architectures, or business logic until every ambiguity is resolved.

grilling: The underlying interrogation engine powering all grilling workflows across the suite.

handoff: Compresses active session state, decisions, and remaining tasks into a portable handoff document for subsequent agent sessions.

teach: Multi-session interactive tutoring workspace that uses the project directory to guide hands-on mastery of advanced concepts.

to-questionnaire: Converts unresolved stakeholder questions into structured Markdown questionnaires tailored for asynchronous responses.

wait-what: Instant re-pitching tool that translates confusing agent explanations into simple, plain English using domain glossary terms.

writing-for-agents: Comprehensive authoring guide for writing documentation, AGENTS.md, and skills that agents interpret with zero drift.

4. Advanced Orchestration and In-Progress Skills

Next-generation multi-agent primitives for continuous software factories:

chief-of-staff: High-level orchestrator coordinating complex cross-functional agent initiatives.

claude-handoff: Dedicated session serialization format tailored for Claude Code subagent handoffs.

loop-me: Autonomous feedback loop runner driving self-correcting agent execution.

setup-ts-deep-modules: Automated scaffolding for TypeScript project structures enforcing deep module boundaries.

software-factory: End-to-end autonomous factory executing tickets from backlog to verified pull request.

writing-beats: Narrative cadence framework for breaking complex technical documentation into digestible beats.

writing-fragments: Modular documentation fragments optimized for targeted agent context injection.

writing-shape: Structural blueprinting skill for shaping technical articles and architectural proposals.

5. Developer Tooling and Guardrails

Operational utilities and safety mechanisms for local agent workspaces:

git-guardrails-claude-code: Pre-commit safety hooks preventing accidental commits of sensitive keys or destructive git actions.

migrate-to-shoehorn: Automated migration assistant for transitioning legacy agent workflows to modern skill specs.

scaffold-exercises: Interactive coding exercise generator for developer onboarding and team skill building.

setup-pre-commit: Automated pre-commit hook configuration for linting, type-checking, and formatting.

Core Architectural Principles

Matt Pocock's skills adhere to four foundational software design principles:

Principle 1: Deep Modules Over Shallow Interfaces

A deep module hides immense internal complexity behind a compact, intuitive interface. Shallow modules add cognitive overhead without reducing complexity. The codebase-design skill actively surveys your project to replace sprawling micro-services with unified, deep modules testable at public seams.

Principle 2: Ubiquitous Domain Vocabulary

Agents waste reasoning tokens guessing project concepts. By enforcing GLOSSARY.md at the repository root, developers and AI agents establish a shared domain language. Concision increases context density, eliminates hallucinated terms, and accelerates reasoning.

Principle 3: Seam-Based Test-Driven Development

Testing internal implementation details makes refactoring brittle. The tdd skill mandates testing exclusively at public seams. When internal logic changes, tests remain green as long as public contract behavior holds.

Principle 4: Two-Axis Parallel Code Reviews

Human reviewers often conflate style conventions with functional correctness. The code-review skill runs two parallel, unpolluted subagents: one auditing against coding standards and Fowler code smells, and the other rigorously verifying that the diff satisfies the originating specification.

Technical Specifications and Compatibility Matrix

Specification	Details
Repository	github.com/mattpocock/skills
Author	Matt Pocock (Total TypeScript, AI Hero)
Total Skills	39 production ready skills
License	MIT Open Source License
Agent Support	Claude Code, OpenAI Codex, GitHub Copilot, Gemini CLI, Cursor, Antigravity
Distribution	Claude Plugin, Codex Marketplace, Copilot Plugin, npx skills CLI
Primary Tag	code / agent
Unlock Price	10 Redit Gold

Official Resources and Links

Official Documentation: https://aihero.dev/skills

GitHub Repository: https://github.com/mattpocock/skills

AI Hero Newsletter: https://www.aihero.dev/s/skills-newsletter

Total TypeScript: https://www.totaltypescript.com

Skills for Real Engineers Matt Pocock AIHero

Interactive Terminal Locked

Overview and Core Capabilities

Comprehensive Breakdown of the 39 Specialized Skills

1. Engineering User-Invoked Skills

2. Engineering Model-Invoked Skills

3. Productivity and Communication Suite

4. Advanced Orchestration and In-Progress Skills

5. Developer Tooling and Guardrails

Core Architectural Principles

Principle 1: Deep Modules Over Shallow Interfaces

Principle 2: Ubiquitous Domain Vocabulary

Principle 3: Seam-Based Test-Driven Development

Principle 4: Two-Axis Parallel Code Reviews

Technical Specifications and Compatibility Matrix

Official Resources and Links

How to Use & Getting Started

Installation Runbook & Walkthrough Locked

Unlock full access to Skills for Real Engineers Matt Pocock AIHero

More in agent

Frontend Developer

Awesome GPT Image

Code Reviewer