mattpocock-skills
claudeMatt Pocock's agent skills for real engineering (grilling, spec/ticket flows, TDD, code review, domain modelling and more).
39 skills
A role-shaped subagent fleet: agents, shared skills, the board and its hooks, and workflows.
Plugin coder-fleet from marketplace rzem in rzem-ai/coder-fleet. Commands read from the repo's manifest — yesterday.
This plugin runs code on your machine — its hooks fire automatically on events like tool use and it starts local MCP servers. The exact commands are listed below; read them before installing.
compoundCapture what a finished unit of work taught you into the one place it will be read again - a project convention into .claude/rules/, a repeatable procedure into a skill, a durable decision into the memory of the agent that needs it - and throw away everything that fails the three tests. The discipli
glossaryThe fleet's shared vocabulary - the agreed meaning of Initiative, Project, Milestone, Issue, Sub-issue, Task, Spec, Phase, Round, Session, Lead, Subagent, Teammate, Handoff, Run article, Gate, Board, Next, Human queue, Eval, Intermittent failure and Sprite, and what each maps to on the board, in Cla
handoffThe handoff format every fleet subagent ends its final message with - the four fixed headings Done, Not done, Unverified and Decisions needed, and the typed Blocker and Propose item lines that the SubagentStop hook parses. Preloaded into every agent; not discovered.
help-boardsHow the board works - its projects and issues, the meaning of the six columns (to do, next, in progress, blocked, blocked by human, done), which columns are written by hooks and which a human-facing assistant writes itself, the checkout's focus (`task_focus`, `/work`) that tells the hooks which issu
humanizeReviews and edits copy to remove AI-generated patterns and make text sound natural. Use when editing drafts, reviewing copy, "humanize this", "make it less AI", "sounds too AI", "remove AI patterns", "edit my copy", "this sounds robotic", or when text feels machine-generated.
lead-proceduresHow the lead carries out each step of its delegation policy - taking an order, the card's comments and ticks, briefing and spawning, the challenge gate and a chosen design, building a change itself, review and escalation, the editors, merging handoffs, shipping, and worktrees and memory. Preloaded i
loopingThe procedure for work that runs the same suite more than once and compares outcomes against a recorded baseline - what counts as a meaningful mutation, why a crashed test process is a kill and not a survival, why some mutants are equivalent and deserve to be named as such rather than chased, and wh
migration-checklistThe twenty checks run over every agent body and every skill's frontmatter, and a pass over their prompt text, when a new model ships, when a frontmatter field changes upstream, or before any change to agent definitions merges. Each check says what to look for and what its failure looks like, and mos
run-articleHow to write a run article - the readable account of one run that the handoff cannot carry, covering what was tried and abandoned, what the constraint turned out to be, what surprised you and what to do differently. Articles live at docs/runs/ in the main checkout, local and unversioned, and the pat
/coder-fleet:agentsList the fleet agents and whether each is enabled for this project, switch one off or on with `disable <agent>` or `enable <agent>`, or show and set the build or harden phase with `phase`
/coder-fleet:askAsk the human every open question on the board with AskUserQuestion - each unticked Actions for Human entry on any card short of Done - and record each answer on its card, moving no column and starting no build
/coder-fleet:boardStart this session's board web UI, or stop it with `stop`
/coder-fleet:drainFinish the cards already In Progress that the human ordered, skipping the rest, then work the Next column top first, card after card, without stopping to ask the human - each card built to a pull request ready to merge, or parked with its question on the card - and end with one line saying what drai
/coder-fleet:flushSay that nothing lands on the default branch, where the board and memory never go, and push the state branch
/coder-fleet:initInitialise the current project for the fleet - settings, AGENTS.md skeleton, glossary rule, spec directory, then a guided fill of every placeholder
/coder-fleet:kickoffPreflight the fleet in this project - plugin, agents, settings, skeleton - check or set up the board and state its conventions, then take the first idea and start the spec pipeline on it
/coder-fleet:prune-worktreesRemove agent worktrees whose work has been adopted - unlocked, clean, merged into the default branch, and under .claude/worktrees/ - and report every one kept and why
/coder-fleet:springcleanClear every Done card to completed when no open card depends on it and none of its child cards is open, in one board commit and without asking, then report each card held and by which open cards
/coder-fleet:sweepReview what merged to the default branch since the last sweep for drift - one helper written twice, one pattern done two ways, dead code, a convention held by habit and no check - and file each finding as a To do card that proposes the check that stops it coming back
/coder-fleet:triageGo through the cards In Progress, Blocked and Blocked by human - close the In Progress ones that are fully ticked and merged, report where the rest stand, and ask the human multiple-choice questions to unblock the blocked cards and answer the open Actions for Human, recording each answer on its card
/coder-fleet:workFocus this checkout on one board item, so the hooks move that item as agents start and stop - the human's hand on what task_focus does for the lead
coderImplements one item or sub-issue the human ordered, from the lead's brief and the board card - a test only where the change needs one, small commits - and reports what changed and what is unverified. Use when the human has ordered the work and the card carries its acceptance criteria.
fleet-stewardKeeps the fleet's definitions current. Runs weekly and unattended, watching for model and tooling changes, running the migration checklist over every agent body, running the evals, and auditing installed plugins. It files and proposes, and never merges.
leadRoutes and gates the fleet. Builds what the human orders from the board card, picks which agent gets which job, writes its brief, holds the escalation policy, and merges the handoffs that come back. Set via `agent` in project settings, not spawned.
researcherReads widely across the web and local files, and returns a synthesis in which every claim carries a source. Use when answering a question needs more reading than an expensive agent should spend its context on.
reviewerChecks a change and returns a verdict with ranked findings. Reviews the diff and runs the declared gates on every spawn, and in the modes the brief names also tries to break the change with mutations or uses the running product against the card's acceptance criteria. Never edits the project. Use aft
scoutFinds where things are and how they work in a codebase and returns paths, line numbers and quoted excerpts - never opinions. Use before an expensive agent starts reading, and for any "where is X" or "how does Y work" question.
scripterImplements one small, well-scoped scripting item or sub-issue the human ordered, from the lead's brief and the board card - a test only where the change needs one, small commits - and reports what changed and what is unverified. Use for scripts, glue and tooling the human has ordered; use coder inst
spec-editorChallenges a spec draft adversarially, once, after spec-writer's interview revision and before the human edits it, writing each challenge into the draft's Challenges section with inline markers and never rewriting its prose. Spawned where AGENTS.md records opus for the spec editor.
spec-writerBuilds the interview brief for a brain dump or a board item and drafts a spec - problem, non-goals, acceptance criteria, open questions - for the human to edit; the lead puts the questions to the human in its own session. Use only for an unshaped idea in a project whose AGENTS.md has no `Requirement
tech-editorEdits a tech-writer document in place, once, after tech-writer hands off and before the document ships, checking each claim against its sources and adding none; a claim it cannot support goes in its handoff, never the prose. Spawned where AGENTS.md records opus for the tech editor.
tech-writerWrites READMEs, ADRs, runbooks, internal docs and blog drafts from material that already exists - a spec, a board card, a diff or the code itself. Use once the work is decided and someone outside the session needs to read about it.
ui-designerTurns an approved spec into screens, flows and self-contained HTML prototypes, arguing for one direction and showing two alternatives. Use after a spec is approved and before a coder builds anything.
boardruns: ${CLAUDE_PLUGIN_ROOT}/board/board.sh mcp
"${CLAUDE_PLUGIN_ROOT}/hooks/board-focus-clear.sh"
"${CLAUDE_PLUGIN_ROOT}/hooks/board-binary.sh"
"${CLAUDE_PLUGIN_ROOT}/hooks/memory-state.sh" start
"${CLAUDE_PLUGIN_ROOT}/hooks/main-fast-forward.sh"
"${CLAUDE_PLUGIN_ROOT}/hooks/board-merged-cards.sh"
"${CLAUDE_PLUGIN_ROOT}/hooks/board-subagent-start.sh"
"${CLAUDE_PLUGIN_ROOT}/hooks/agent-clock.sh"
"${CLAUDE_PLUGIN_ROOT}/hooks/memory-state.sh" start
"${CLAUDE_PLUGIN_ROOT}/hooks/board-subagent-stop.sh"
"${CLAUDE_PLUGIN_ROOT}/hooks/memory-state.sh" stop
"${CLAUDE_PLUGIN_ROOT}/hooks/board-task-completed.sh"
"${CLAUDE_PLUGIN_ROOT}/hooks/board-agent-return.sh"
"${CLAUDE_PLUGIN_ROOT}/hooks/board-release-close.sh"
"${CLAUDE_PLUGIN_ROOT}/hooks/memory-state.sh" post
"${CLAUDE_PLUGIN_ROOT}/hooks/enforce-agent-scope.sh"
"${CLAUDE_PLUGIN_ROOT}/hooks/agent-clock.sh"
"${CLAUDE_PLUGIN_ROOT}/hooks/enforce-disabled-agents.sh"
"${CLAUDE_PLUGIN_ROOT}/hooks/memory-state.sh" stop
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
Official Next.js skills: adopt and optimize Cache Components, adopt Partial Prefetching, and verify runtime behavior against a running dev server.
6 skills
Talk like caveman. Cut filler. Keep technical details explicit.
22 skills · 1 command · 3 agents · hooks
Persistent memory for Claude Code: the next session starts where the last one left off.
29 skills · 1 MCP server · hooks
tasteskill 2.5.1: professional company websites at the figma.com and x.ai level, with one section per screen, real photographs, real product UI, effects kept for tech and gaming pages, and a real-browser checker and finish gate.
1 skill
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.