mattpocock-skills
claudeMatt Pocock's agent skills for real engineering (grilling, spec/ticket flows, TDD, code review, domain modelling and more).
39 skills
Workbench for structured software development with TDD, project planning, and beads integration.
Plugin wb from marketplace gvarela-workbench in gvarela/workbench. Commands read from the repo's manifest — 1 month ago.
This plugin runs code on your machine — its hooks fire automatically on events like tool use. The exact commands are listed below; read them before installing.
create_designWrite docs/plans/<project>/design.md: WHAT to build and WHY — decisions, rationale, scope, success criteria, risks — from validated research. Use when research.md is complete and the user asks to design, decide the approach, or write the design doc, or to formalize a decision explore_design recorded
create_handoffWrite a handoff document that carries this session's working context — decisions, discoveries, deviations from plan, beads state, next steps — into a new session, model, or agent. Use when ending or pausing a session, switching models or machines, when asked to hand off, checkpoint, or save context,
create_mockupResearch existing UI patterns and produce an initial mockup for a feature, with clarifying questions for the user. Use when the user asks for a mockup, wireframe, or UI draft for a planned feature; iterate afterwards with mockup-iteration. Takes the project directory and a feature description.
create_product_researchResearch a codebase from the product perspective — features, user flows, user-visible behaviors — into research.md written in product language. Use when the research question is about what the software does for users rather than how the code works, or when a PM-facing write-up is requested. Takes th
create_projectStart a new wb project: create docs/plans/<date>-<name>/ with README, research.md, design.md, and tasks.md skeletons and record git metadata. Use when the user wants to plan, scope, or kick off a feature, refactor, or investigation as a tracked project, or asks to set up plan docs. First stage of th
create_researchResearch how something currently works in this codebase and write docs/plans/<project>/research.md: facts only, file:line references, no recommendations. Spawns parallel locator, analyzer, and pattern-finder agents. Use when a project directory exists and the user asks to research, investigate, docu
create_tasksTurn an approved design.md into docs/plans/<project>/tasks.md: phased, executable tasks with file:line targets and verification, plus the beads epic, milestones, task issues, and dependencies. Use when design is approved and the user asks to plan the implementation, break the work down, create tasks
doc-adherenceUse when about to state what a plan document says (per the design, the plan calls for, research shows, as documented) or when resuming work after context compaction - requires the claim to come from a read of that file in the current context window, not from a summary or memory.
explore_designFacilitated architecture discussion between research and design: explores directions as possibilities, converges only on explicit user approval, and records the decision durably (thoughts/ doc plus a closed Decide: issue). Use when a design has open trade-offs or competing directions, or the user wa
helpReference for the wb workflow and, in a repository with an active plan, where you are and what's next: which documents exist and their status, epic and milestone state, open Q: and Decide: issues, what the next stage needs from you, and what the previous stage left undone. Use when the user asks whe
implementImplement a project's tasks.md by coordinating fresh-context worker agents (one per task, model chosen per task), verifying each with a task-verifier, and escalating verified failures. The default execution path: use when the user asks to implement, build, or execute a planned phase, asks for worker
implement_inlineImplement a project's tasks.md in this session with TDD (red, green, refactor), beads claim and close per task, and a human checkpoint at each phase boundary. Use only when the user asks for the work done inline, in this session, or on the session model rather than by workers (for example a full-Fab
mockup-iterationIterate on UI mockups, capturing keeps/removes/changes with full fidelity. Versions each iteration and maintains decision log.
project-structureUse when writing or editing research.md, design.md, tasks.md, or a thoughts/ doc under docs/plans/, or when deciding which document a piece of content belongs in: research.md for facts, design.md for decisions and rationale, tasks.md for implementation steps, thoughts/ for explorations.
research-validationValidate a research document (research.md or product-research.md) against the current codebase: file paths exist, code snippets match, behavioral claims hold. Use when the user asks to validate, verify, or fact-check research, or asks whether research is still accurate after the code changed. Takes
resume_handoffResume work from a handoff document written by create_handoff: reload the context, decisions, discoveries, and next steps, then continue from the ready work in beads. Use at the start of a session when the user points at a handoff file or asks to pick up, resume, or continue prior work. Takes the ha
review-prepInteractive pair-programming walkthrough of a diff using tmux and nvim: opens each changed file in an nvim pane and takes questions as you go. Use when the user wants to be walked through what changed, to understand a diff by reading it together, or to prep a PR that way; not for an automated review
status-syncUse when ending a session, wrapping up work, or completing a phase - monitors beads status and reminds about committing beads state and closing finished phase milestones.
tdd-disciplineUse when implementing features, fixing bugs, or writing ANY production code in this session - including quick ad-hoc fixes outside the workflow commands (coordinated workers preload this skill; solo edits must trigger it). Enforces RED-GREEN-REFACTOR cycle where tests must fail before writing implem
update_statusReconcile a wb plan's documentation status from beads — the sole writer of tasks.md frontmatter progress fields (status, current_phase, completed_tasks) and the status lines in research.md and design.md. Use when a phase or task set completes, when asked to update, sync, or reconcile plan status or
validate_executionValidate that an implemented plan matches design.md and tasks.md: run the automated checks, compare the implementation to the requirements, and produce a pass/fail report with gaps. Use after a phase or project is implemented when the user asks to validate, verify, or review whether the plan was don
validate_projectCheck that a docs/plans/<project>/ directory follows the wb workflow: required files, frontmatter, status transitions, beads tracking, and stage ordering. Use when plan documents look inconsistent, when asked to validate or audit a project's docs, or before resuming an old project. Takes the project
verification-before-completionUse when about to claim work is complete, fixed, passing, or done - including after ad-hoc fixes and one-off commands, not just workflow phases. Requires running verification commands and confirming output before making any success claims. Evidence before assertions.
codebase-analyzerAnalyzes codebase implementation details. Finds how specific components work, traces data flow, and documents technical details with precise file:line references.
codebase-locatorFinds specific components and files in the codebase. Specializes in locating where features are implemented, which files contain certain functionality, and mapping the codebase structure.
pattern-finderFinds similar patterns and implementations in the codebase. Identifies conventions, recurring patterns, and examples that can guide new implementations.
research-validatorValidates research documents by checking file paths exist, code snippets match actual files, and behavioral claims are accurate against the codebase. Returns structured PASS/FAIL report.
task-verifierVerifies task completion by running tests, checking scope adherence, and validating implementation against requirements. Returns structured pass/fail report.
task-workerFocused implementation worker for exactly one beads task under coordinated execution. Implements the single task it is given with strict TDD, claims and closes its beads issue, and returns a structured completion report. Spawned by /wb:implement with a per-task model override.
${CLAUDE_PLUGIN_ROOT}/hooks/wb-prime.sh
${CLAUDE_PLUGIN_ROOT}/hooks/wb-prime.sh
${CLAUDE_PLUGIN_ROOT}/scripts/lint-hook
${CLAUDE_PLUGIN_ROOT}/scripts/lint-hook
${CLAUDE_PLUGIN_ROOT}/hooks/beads-drift-check.sh
Matt Pocock's agent skills for real engineering (grilling, spec/ticket flows, TDD, code review, domain modelling and more).
39 skills
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques
15 skills · hooks
Up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your coding agent's context.
1 MCP server
Backend API design, GraphQL architecture, workflow orchestration with Temporal, and test-driven backend development
9 skills · 1 command · 8 agents
Search and retrieve documents from local markdown files.
2 skills
AI-supervised issue tracker for coding workflows
1 skill · 1 agent · hooks
This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.