wizard
๐ฏSkillfrom mattpocock/skills
Installation
npx vibeindex add mattpocock/skills --skill wizardnpx skills add mattpocock/skills --skill wizard~/.claude/skills/wizard/SKILL.md
A collection of composable, model-agnostic agent skills by Matt Pocock that address common coding-agent failure modes: misalignment, verbosity, broken code, and architectural decay. Includes grilling sessions, TDD loops, domain modeling, and codebase architecture analysis.
Overview
A curated set of small, composable agent skills built on decades of software engineering experience. Rather than owning the entire development process like monolithic frameworks (GSD, BMAD, Spec-Kit), these skills are designed to be easy to adapt and hack on. They work with any model and install in 30 seconds via the skills.sh CLI. The collection targets four common coding-agent failure modes: misalignment between developer intent and agent output, excessive verbosity, code that does not work, and accumulating architectural complexity.
Key Features
- Grilling sessions (
/grill-me,/grill-with-docs): structured interview loops that force alignment between you and the agent before any code is written./grill-with-docsalso builds a shared domain language inCONTEXT.mdand records decisions as ADRs. - Test-driven development (
/tdd): a red-green-refactor loop that gives the agent continuous feedback on whether the code actually works, producing far more reliable output. - Bug diagnosis (
/diagnosing-bugs): a disciplined reproduce-minimize-hypothesize-instrument-fix-regression-test loop for hard bugs and performance regressions. - Architecture improvement (
/improve-codebase-architecture): scans for deepening opportunities and generates a visual HTML report, then grills you through whichever improvement you pick. - Planning and triage (
/to-prd,/to-issues,/triage): converts conversations into PRDs, breaks plans into vertical-slice issues, and moves issues through a triage state machine. - Handoff and teaching (
/handoff,/teach): compacts a conversation into a handoff doc for another agent, or teaches a concept over multiple sessions using the current directory as workspace.
Who is this for?
- Software engineers who use Claude Code, Codex, or other coding agents and want more reliable, higher-quality output from their AI workflows.
- Developers frustrated by agents that produce misaligned code, overly verbose responses, or unmaintainable architecture.
- Teams looking for lightweight, composable practices they can adopt incrementally without committing to a heavyweight framework.
- Anyone who values software engineering fundamentals like TDD, domain-driven design, and iterative refinement and wants to bring those disciplines into AI-assisted development.
Same repository
mattpocock/skills(58 items)
SKILL.md
More from this repository10
Interviews the user relentlessly about every aspect of a plan or design until reaching shared understanding, resolving each branch of the decision tree one by one. Provides recommended answers for each question and explores the codebase when answers can be found there.
A skill that runs a grilling session to challenge your plan against an existing domain model, sharpen terminology by building a shared language in CONTEXT.md, and document architectural decisions in ADRs before coding begins.
An agent skill from Matt Pocock's engineering toolkit that helps improve your codebase architecture. Part of a composable skill set designed to fix common AI coding agent failure modes through structured, experience-based workflows.
Enforces test-driven development with a strict red-green-refactor loop using vertical slices (one test then one implementation at a time), emphasizing behavior verification through public interfaces rather than implementation-coupled testing.
Initial setup skill for Matt Pocock's engineering skills collection that configures your backlog manager, triage labels, and documentation paths for AI coding agents.
A model-invoked productivity skill that interviews users relentlessly about a plan or design until every branch of the decision tree is resolved. It serves as the reusable interview loop behind the grill-me and grill-with-docs skills.
Composable, model-agnostic agent skills for real engineering workflows, featuring /grill-me for requirement alignment, /prototype for rapid iteration, and task management tools designed to fix common failure modes in AI coding agents.
A model-invoked engineering skill that provides shared discipline and vocabulary for designing deep modules: maximizing behavior behind small interfaces, placing code at clean seams, and ensuring testability through public APIs.
A model-invoked skill that guides coding agents through a disciplined diagnosis loop for hard bugs: reproduce, minimise, hypothesise, instrument, fix, and regression-test.
A collection of small, composable engineering skills designed for real development workflows, featuring grilling sessions for requirement alignment, issue tracking integration, and task triage with customizable labels.