mattpocock-skills
claudeMatt Pocock's agent skills for real engineering (grilling, spec/ticket flows, TDD, code review, domain modelling and more).
39 skills
itGenius's public Claude Code skills: the full dev process (dev-brainstorming, dev-writing-plans, dev-quick-build, dev-auto-build, dev-qa-verify, dev-debugging, dev-deploy-safely), multi-model review (review-swarm over Codex, Gemini, GLM, and an isolated Claude pass; plus review-debate), session and repo hygiene, VM operations, executive summaries, checked PDFs, AI-copy humanizing, a daily morning-brief HUD, and checked write layers for Google Chat and Asana.
Plugin skills-library from marketplace skills-library in itgenius-au/skills-library. Commands read from the repo's manifest — 9 days ago.
ai-debateALIAS - renamed to review-debate. Forwards to `review-debate`. Kept so `/ai-debate` and older references still resolve. Triggers on: ai debate, debate this, three-model review, consensus review, multi-model debate, 3-way review.
asana-updateEvery Asana write: a comment (new or edited), a task create or close, a field change, a description append, a file attachment or a time entry. Use it BEFORE any Asana write, even unprompted. Never write your own script that calls the Asana API. Triggers: 'update Asana', 'update the Asana status', 'p
claude-vm-opsOperate and safely fix a Linux VM that runs Claude Code Remote Control servers, laid out as in docs/claude-code-vm-setup-spec.md (one systemd unit and one Linux user per Claude account). Covers a slow or hung VM, high load, a full disk, a lost or expiring Claude sign-in, and servers left on an old C
cleanup-gitALIAS - renamed to repo-cleanup. Forwards to `repo-cleanup`. Kept so `/cleanup-git` and older references still resolve. Triggers on: cleanup-git, clean up git, tidy branches, git is a mess, audit my branches.
codex-reviewALIAS - renamed to review-with-codex. Forwards to `review-with-codex`. Kept so `/codex-review` and older references still resolve. Triggers on: codex review, codex check, codex audit, second opinion from codex, ask codex.
copy-humanizerRemove AI-generated patterns and restore natural human voice to marketing copy, blog posts, and email content. Triggers on: humanize, copy humanizer, de-ai, remove ai patterns, make it human, sounds like ai, too robotic, ai voice.
deploy-safelyALIAS - renamed to dev-deploy-safely. Forwards to `dev-deploy-safely`. Kept so `/deploy-safely` and older references still resolve. Triggers on: deploy-safely, deploy safely, safe deploy, deploy gates, pre-deploy checks.
dev-auto-buildAutonomously execute a plan/runbook end to end using a rigorous multi-model method: an AI build agent (e.g. Codex) builds per phase, Claude verifies independently, review swarm (four independent models: Codex, Gemini, GLM and a deep-reasoning Claude pass) cycles findings back to the build agent unti
dev-brainstormingTurn a vague idea into a concrete, approved design or spec through collaborative dialogue. Classifies the request first (spike / bounded / architectural) so the ceremony fits the work, holds a hard no-build-before-approval gate, runs divergent options -> design -> spec grounded in real research, can
dev-debuggingThe house debugging discipline for a bug or unexpected behaviour: reproduce it on demand, form one testable hypothesis, isolate the cause with evidence (measure live state, never reason from the docs or the recipe), fix the root cause not the symptom, then verify the reproduction passes and lock it
dev-deploy-safelyUse before production, shared staging, or preview deploys. Verifies environment class, branch freshness, worktree state, tested commits, deploy manifest, smoke checks, and rollback path. Triggers on: dev deploy safely, dev-deploy-safely, deploy safely, deploy-safely, safe deploy, deploy gates, pre-d
dev-qa-verifyThe verification-before-completion gate: prove a change works on the real surface before anything is called done. Runs an interactive browser pass (interact, do not just look) and triages any defect by severity - a build bug loops back to the build skill, a plan-level miss routes to dev-writing-plan
dev-quick-buildBuild and ship ONE small, well-scoped feature end to end in a single session - worktree isolation, dated runbook doc kept updated throughout, a quick review-swarm plan review before any code (four models, about 10 minutes), TDD, full local gates, adversarial multi-agent review, dev-deploy-safely pro
dev-update-envAudit and update the CLI tools, MCP servers, and system packages on a macOS or Linux dev machine. Detects what is installed (npm, uv, pipx, Homebrew, apt, dnf), compares installed vs latest versions, presents one summary table, and runs updates only after confirmation. Triggers on: update environmen
dev-writing-plansTurn a spec into a granular, dependency-ordered implementation plan with a file-structure map, right-sized tasks, verification criteria, risk/rollback flags, and checkpoints - then harden it with a review-swarm-until-consensus loop before any build. Use when the user says 'write plan', 'write a plan
document-pdfMake a checked PDF from one Markdown source: reports, memos, letters, statements, translations, agreements, deeds, resolutions, claim letters and evidence bundles. Builds on macOS or Linux, runs a verify gate (facts present, no placeholders, page size, a look at every page), and hands over a draft,
exec-summaryTurn complex content into a short plain-language executive summary in Simplified Technical English: key points, numbered decisions each with a recommendation, next steps, then the bottom line and a 'go' reply line. Use it WHENEVER the user asks for a summary or recap, asks where the work stands, ask
fable-reviewALIAS - renamed to review-with-claude. Forwards to `review-with-claude`. Kept so `/fable-review` and older references still resolve. Triggers on: fable review, fable-review, review with fable, ask fable, fable check, fable audit, fable opinion, isolated claude review, second claude opinion.
gemini-reviewALIAS - renamed to review-with-gemini. Forwards to `review-with-gemini`. Kept so `/gemini-review` and older references still resolve. Triggers on: gemini review, gemini check, gemini audit, ask gemini, gemini opinion, google review.
google-chat-updatePost a Google Chat message consistently, under the right identity: as yourself (admin DWD users only) when the user asked for the post, as your bot when the agent posts on its own or you have no DWD. Use this skill BEFORE any Chat post you are asked to send, draft-and-send, or reply on someone's beh
machine-moveLive multi-machine sync (the daily driver) OR one-time new-machine setup. SYNC pushes all git state online, pulls the other machine's work, prunes merged branches, keeps tooling current and makes local-only data durable; its dry run is also the pre-flight audit before you leave a machine. SETUP boot
morning-briefUse when the user runs /morning-brief or asks for their morning brief, daily briefing, "what's on today", "brief me", "catch me up on my day", or a rundown of calendar + tasks + email + chats. Pulls today's Google Calendar schedule first, then Asana, unread Gmail, and Google Chat; triages signal fro
plugin-fresh-install-testTest what a teammate on a brand-new machine actually gets from a Claude Code plugin (skills, SessionStart context, PreToolUse guards) - with a throwaway config dir, optional logging stubs for risky commands, and evidence graded from events instead of the model's self-report. Works for any plugin. Us
quad-reviewALIAS - renamed to review-swarm. Forwards to `review-swarm`. Kept so `/quad-review` and older references still resolve. Triggers on: quad review, quad-review, quad check, four-model review, 4-model review, panel review, Codex Gemini GLM Fable.
repo-cleanupSafely audit and clean up accumulated git state across a repo - orphan branches, old worktrees, "wip" commits, diverged main, stale remote-tracking branches, local-only work that needs preserving. Use whenever the user says "repo cleanup", "clean up this repo", "cleanup git", "clean up git", "tidy t
repo-context-cleanupAudit and clean up the standing context Claude Code carries into every session - CLAUDE.md files and the auto-memory store - so a fresh session orients fast and the always-loaded context stays lean and accurate. Three modes: `claude-md` scores CLAUDE.md files against a quality rubric and applies tar
repo-health-sweepGreen-lane loop that sweeps your code repos, runs each repo's OWN declared gates (test/lint/typecheck), and reports a health map. In --fix mode it hands each fixable failing repo to dev-auto-build to produce a verified push-ready fix branch. Nothing is pushed, merged, or deployed. Triggers on: repo
review-debateOrchestrate multi-round debates between Claude, Gemini, and Codex for higher-quality reviews. Three rounds: independent review, targeted challenge on disputes, and synthesis with self-challenge. Triggers on: review debate, review-debate, review with debate, ai debate, debate this, three-model review
review-swarmRun a rigorous review swarm - four models (OpenAI Codex + Google Gemini + z.ai GLM + a deep-reasoning Claude Fable pass) over the same subject in one parallel pass, then synthesize one deduplicated, severity-ranked, consensus-tagged findings list. Three are INDEPENDENT external models; the fourth (F
review-with-claudeRun a second, isolated Claude Code instance (nicknamed \"Fable\" by this skill) headless and read-only via `claude -p`, as a deep-reasoning second-opinion reviewer for code, decisions, plans, and builds. You choose the model and reasoning effort it runs at - a Fable-tier or other deep-reasoning mode
review-with-codexUse OpenAI Codex CLI as a second-opinion reviewer for code, decisions, plans, and builds. Runs Codex headlessly through a bundled safety wrapper to get independent analysis from a different AI model, then synthesizes findings. Triggers on: codex review, second opinion, codex check, review with codex
review-with-geminiUse Google Gemini CLI as a second-opinion reviewer for code, decisions, plans, and builds. Runs Gemini headlessly and read-only through a bundled safety wrapper to get independent analysis from a different AI model, then synthesizes findings. Triggers on: gemini review, gemini check, review with gem
review-with-glmUse z.ai's GLM models (its current flagship by default) as a second-opinion reviewer AND a general coding engine, via the z.ai HTTP API. A third independent model alongside review-with-codex and review-with-gemini - for code/decision/plan/build/codebase review, and for offloading coding tasks to GLM
session-dspSquash the per-turn `wip` commits from this session into one clean commit and push. Worktree-aware: skips the attribution check on a worktree branch (each worktree = one session, no collisions possible). Pairs with the bundled wip-commit hook (scripts/auto-commit.sh). Triggers on: session dsp, dsp,
session-recoveryBring offline Claude Code background sessions back and clear the backlog: find each session that claude.ai/code shows as disconnected or offline, wake it on its own thread (never a copy), have it review itself read-only in a fixed readable format, merge duplicates under one owner, prepare each owner
session-status-sweepReport where every other live Claude Code session stands, read-only: the ask, repo/worktree/branch, what is done, current state (quoting any question waiting on the user), and last activity, then list the sessions waiting on the user and the ones stuck or abandoned. Works on a local Mac, on a shared
session-wrapEnd or checkpoint a working session cleanly with four distinct intents - done (complete: integrate and close), pause (staying in this window), pick up (window closing, compacting or context full: hand off a self-sufficient resume prompt), or skill sweep (deep review of the session's skill and doc le
triple-reviewALIAS - retired; forwards to review-swarm. Kept so `/triple-review` and older references still resolve. Triggers on: triple review, triple-review, triple check, triple-check this, three-model review, 3-model review, review with all three, Codex Gemini GLM, panel review.
zai-reviewALIAS - renamed to review-with-glm. Forwards to `review-with-glm`. Kept so `/zai-review` and older references still resolve. Triggers on: zai review, z.ai review, glm review, ask glm, glm check, glm audit, glm opinion, glm coding, third opinion.
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
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
Open-source AI job search that runs locally in your AI coding CLI: scan job portals, evaluate listings into a structured A-H report with a 1-5 score, tailor your CV, track applications
5 skills · 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.