upscaler
uwg · Kit
Holds a set of image upscaling models gathered from the OpenModelDB community, for people who want to enlarge low-resolution pictures with AI on their own comp…
New
Overview
Matt Pocock’s free, MIT licensed agent skills for everyday engineering: interview-style planning, specs and tickets, test-first building, code review and bug diagnosis. The setup notes cover Claude Code, Codex, GitHub Copilot, Gemini CLI and skills.sh.
Skills For Real Engineers is an open-source project on GitHub in the Developer category, listed by Mindket. It is free. This is version 2, last updated on . Install it from GitHub as its README says. Get Mindket’s setup notes on your Mindket account, as a zip.
At a glance
- Runs in
- Claude Code: a plugin from the official marketplace that updates itself
- Codex: a plugin from the maker’s own marketplace that updates at startup
- GitHub Copilot: the CLI, which updates itself after a one-time opt-in, and VS Code, which updates daily
- Gemini CLI: two install commands, updated by hand
- Start with
claude plugin install mattpocock-skills@claude-plugins-official
Platform
Claude Code
Codex
GitHub Copilot
VS Code
Gemini CLI
Cursor
And 2 more in Where to use it
About
What it is
Skills For Real Engineers is Matt Pocock’s set of agent skills for everyday engineering work. Each skill is a folder with a SKILL.md that your coding agent follows when you type its name or when the task fits. His README calls them small, easy to adapt and composable, and says they work with any model. It sets them against approaches that own the whole process. Those, it says, take away your control and make bugs in the process hard to resolve. The skills aim at four failures: the agent builds something you did not mean, it talks too much, the code does not work, and the codebase turns into a ball of mud.
- You install the skills and run the setup skill once per repo
- You type /grill-with-docs and the agent interviews you
- /to-spec and /to-tickets turn that into a spec and tickets
- /implement builds one ticket test first, then reviews it
Who it suits
Developers who already work with a coding agent such as Claude Code or Codex and want a plan, tickets and tests behind each change. The README calls them skills for real engineering, not vibe coding. Skip it if you want a framework that runs the whole process for you.
| Industry | Who uses it | How they use it | Example |
|---|---|---|---|
| Software teams | Developers who build with a coding agent | Run an interview first, then turn it into a spec and tickets before any code | A developer describes a new feature, answers the agent’s questions and gets a spec and tickets in the issue tracker. |
| Solo projects | One person with no team tracker | Keep issues as Markdown files in the repo, with no remote needed | A solo developer picks local files during setup and works the tickets blockers first. |
| Open source maintenance | Maintainers who receive bug reports and requests | Sort incoming issues through labelled triage roles | A maintainer asks triage what needs attention, approves its label call for each issue, and gets a brief for the agent on the ones that are ready. |
| Plans and writing | Anyone outside a repository | Run the stateless interview on a plan, a design or a piece of writing | A writer starts grill-me on an essay outline and answers until every branch is settled. |
Where to use it
The README gives an install route for each agent below, and the kit has a page for each.
- Claude Code: a plugin from the official marketplace that updates itself
- Codex: a plugin from the maker’s own marketplace that updates at startup
- GitHub Copilot: the CLI, which updates itself after a one-time opt-in, and VS Code, which updates daily
- Gemini CLI: two install commands, updated by hand
- Any other agent, such as Cursor, OpenCode, Devin, Windsurf, Amp or Pi: the skills.sh installer, updated by hand
How to set it up
Get it saves a zip of Mindket’s setup notes in your downloads folder, with a page for each route above. The README says to pick one route per agent, because installing both the plugin and skills.sh gives you every skill twice.
| Need | Detail |
|---|---|
| A coding agent | Claude Code, Codex, GitHub Copilot, Gemini CLI or another agent that skills.sh supports |
| Node.js | 22.20 or newer, only for the skills.sh route |
| An issue tracker | GitHub with the gh tool, GitLab with glab, local Markdown files, or one you describe |
| A repository | For GitHub or GitLab the setup skill needs a reachable repository. If no remote points at that host, it asks for the URL and adds it (git init if needed). Local Markdown needs no remote, and grill-me needs no repository |
- Install the skills with the route for your agent.
- Run the setup skill once in each repo you work on.
- Type a skill and give it your idea.
For Claude Code, the install is one command.
claude plugin install mattpocock-skills@claude-plugins-official
For most other agents, the skills.sh installer takes your agent’s name. When it asks which skills to take, include setup-matt-pocock-skills.
npx skills@latest add mattpocock/skills -a <agent>
Then, in your agent, inside the repo, run the setup skill. It asks where your issues live and, when triage is installed, whether to keep the default triage labels. It puts the domain docs in one GLOSSARY.md and docs/adr/ unless it sees a monorepo.
/setup-matt-pocock-skills
How to use it
Type a skill as a slash command, or in Codex as $ and the skill name. The README calls grill-me and grill-with-docs its most popular skills and says to use them every time you want to make a change. grill-with-docs interviews you in rounds. It writes terms you settle into GLOSSARY.md and hard-to-reverse decisions into ADRs under docs/adr as it goes.
/grill-with-docs
When the interview ends, keep the same conversation and hand it to the next skill. If you are not sure which to type, ask-matt routes you.
/ask-matt
Example
You want to add a feature to an app. You type /grill-with-docs and describe it. The agent asks about the flow and the edge cases, and you answer. Then you run /to-spec and /to-tickets in the same conversation.
/to-spec /to-tickets
Clear the context and run /implement once per ticket, each starting fresh from its ticket.
/implement
to-spec turns the conversation into a spec on your tracker. to-tickets splits it into tickets that each name what blocks them. implement builds one ticket, driving tdd and closing with code-review.
For newcomers and experts
If you are new. A skill is a text file of instructions your agent follows. Install one route for your agent, run the setup skill once per repo, and start with grill-me or grill-with-docs. grill-me needs no repo and writes no files. grill-with-docs leaves files in your repo. Start grill-me in a fresh conversation, and answer the questions yourself. The maker’s docs say a session where you only agree gives you a plan the agent wrote and you nodded at.
For experts. The skills split on who can start them. User-invoked skills run only when you type them and orchestrate. Model-invoked skills, such as tdd, diagnosing-bugs and code-review, can be typed or reached for by the agent when the task fits. A user-invoked skill may call model-invoked ones, never another user-invoked one.
grill-with-docs is one line that calls grilling and domain-modeling, so installing it alone gives a skill that does not work. tdd writes tests only at seams you agree before the first test. implement-spec runs implementer subagents in parallel across the ready tickets and lands the work on one integration branch. wayfinder maps work too big for one session as decision tickets, then hands off to to-spec.
Good to know
Matt Pocock made these skills. Mindket only wrote the setup notes.
- What it writes: the setup skill adds docs/agents files and an Agent skills block to CLAUDE.md if it exists, else to AGENTS.md. If neither exists, it asks you which to create. In a repo with a CLAUDE.md that Codex ignores, move the block by hand. grill-with-docs writes GLOSSARY.md and ADRs. to-spec and triage post to your issue tracker
- Your data: the repository describes no server or data collection of its own. The skills are instructions, so whatever your agent reads or runs because of one goes through your agent and your model provider
- skills.sh usage data: its README says the CLI collects anonymous usage data, and DISABLE_TELEMETRY=1 or DO_NOT_TRACK=1 turns it off
- Labels: since version 1.3.2 the setup skill creates missing triage labels on GitHub or GitLab, but it never creates wayfinder’s wayfinder:map and wayfinder:<type> labels. Create those by hand before the first wayfinder run on GitHub
- Contributions: the SCOPE file says new-skill proposals are closed, and CONTRIBUTING.md says pull requests are open to collaborators only
- Testing: Mindket copied every command from the README, the maker’s install notes, the skill files and the skills CLI’s README, and ran none of them
- Removal: the README gives no uninstall steps. The skills CLI has npx skills remove
- License: MIT. The LICENSE file reads “Copyright (c) 2026 Matt Pocock”
README and files
# Skills For Real Engineers: setup notes This kit holds setup notes only. Mindket wrote them. The skills, the plugin and the install routes come from Matt Pocock's repository: https://github.com/mattpocock/skills These notes are not made or endorsed by Matt Pocock. His README calls the repository "Skills For Real Engineers" and says: "My agent skills that I use every day to do real engineering - not vibe coding." No skill file from the repository is copied here. License: MIT. The LICENSE file (https://github.com/mattpocock/skills/blob/main/LICENSE) reads "Copyright (c) 2026 Matt Pocock". The pages these notes came from: the README (https://github.com/mattpocock/skills/blob/main/README.md), the maker's install notes (https://github.com/mattpocock/skills/blob/main/.agents/install-block.md), SCOPE.md (https://github.com/mattpocock/skills/blob/main/SCOPE.md), CONTRIBUTING.md (https://github.com/mattpocock/skills/blob/main/CONTRIBUTING.md), each skill's SKILL.md under https://github.com/mattpocock/skills/tree/main/skills, and the docs pages under https://github.com/mattpocock/skills/tree/main/docs, such as setup-matt-pocock-skills, grill-me and grill-with-docs. The facts about the skills.sh installer come from its own README (https://github.com/vercel-labs/skills). The commands were copied from these pages and were not run by Mindket. ## What the README says it is The skills are "small, easy to adapt, and composable", and "They work with any model." The README sets them against approaches that own the whole process, which, it says, take away your control and make bugs in the process hard to resolve. It names four failures they aim at: the agent did not do what you wanted, the agent is too verbose, the code does not work, and the codebase becomes a ball of mud. ## What installs The skills that install sit in two of the repository's bucket folders, engineering and productivity. `misc/`, `in-progress/` and `deprecated/` are not promoted and do not ship in the plugin. The README splits each folder in two: - User-invoked skills run only when you type them, and their job is to orchestrate. - Model-invoked skills can be typed by you or reached for by the agent when the task fits. They hold the reusable discipline. A user-invoked skill may call model-invoked skills, but never another user-invoked one. Engineering, user-invoked: - ask-matt: you ask it which skill or flow fits your situation. It is a router over the user-invoked skills in this repo. - grill-with-docs: a grilling session that also builds the project's domain model and updates GLOSSARY.md and ADRs. - triage: moves issues through a state machine of triage roles. - improve-codebase-architecture: scans a codebase for deepening opportunities and presents them as a visual HTML report. - setup-matt-pocock-skills: configures a repo for the engineering skills, once per repo. - to-spec: turns the current conversation into a spec and publishes it to the issue tracker. - to-tickets: breaks a plan or spec into tracer-bullet tickets, each declaring its blocking edges. - implement: builds the work a spec or set of tickets describes, driving tdd and closing with code-review. - implement-spec: implements a whole spec on one integration branch, with implementer subagents working the ready tickets. - wayfinder: plans a huge chunk of work, more than one agent session can hold, as a shared map of decision tickets on the issue tracker. - retro: suggests improvements to the coding agent's environment after a session. Engineering, model-invoked: prototype, diagnosing-bugs, research, tdd, domain-modeling, codebase-design, code-review, pr and wizard. Productivity, user-invoked: grill-me, handoff, teach, to-questionnaire and wait-what. Productivity, model-invoked: grilling and writing-for-agents. Source: https://github.com/mattpocock/skills/blob/main/README.md (section "Reference") ## What you need first - A coding agent: Claude Code, Codex, GitHub Copilot (CLI or VS Code), Gemini CLI, or another agent the skills.sh installer supports. - Node.js 22.20 or newer, only for the skills.sh route, which runs `npx skills`. The `skills` package on npm lists it in its engines field (https://registry.npmjs.org/skills/latest). - A place for issues. The setup skill offers GitHub (the `gh` CLI), GitLab (the `glab` CLI), local Markdown files, or another tracker you describe. A GitHub or GitLab tracker needs a reachable repository. If no remote points at that host, the setup skill asks for the URL and adds it (`git init` if needed). ## Install by agent The README says to pick one route per agent, because installing both the plugin and skills.sh gives you every skill twice. - setup/claude-code.md - setup/codex.md - setup/github-copilot.md - setup/gemini-cli.md - setup/other-agents.md: Cursor, OpenCode, Devin, Windsurf, Amp, Pi and others, through skills.sh - setup/first-run.md: running /setup-matt-pocock-skills in a repo, and the flow the skills follow ## Updating and removing Each install route has its own update rule, listed in the file for that agent. The README gives no removal steps. For the skills.sh route, the skills CLI has `npx skills remove` (https://github.com/vercel-labs/skills). ## Data The pages read for these notes (the README, the install notes, SCOPE.md and CONTRIBUTING.md) describe no server or data collection by the skills themselves. The skills are instructions your agent follows, so whatever a skill has your agent read goes to your model provider. The skills.sh installer is a separate tool and says it collects anonymous usage data. See setup/other-agents.md. ## Following changes The README links a newsletter for changes to the skills and new ones: https://www.aihero.dev/s/skills-newsletter ## Contributing CONTRIBUTING.md says pull requests are open to collaborators only. SCOPE.md says proposals and contributions for new skills are closed. The skills.sh route gives you editable copies for your own changes, as SCOPE.md says.
README.md 6 KB setup/claude-code.md 2 KB setup/codex.md 2 KB setup/first-run.md 5 KB setup/gemini-cli.md 1 KB setup/github-copilot.md 2 KB setup/other-agents.md 3 KB
What’s new
Updated by Mindket to the current version of this kit.
Reviews
What people say
Posts and comments that name this project on Hacker News, Bluesky, X, Reddit, Facebook and YouTube, found by a weekly search and on Trendshift. Mindket doesn’t write or check them. Open one to read it where it was posted.
ronpotokShow HN: Interview your analysts instead of auto-generating a semantic layerI’ve spent the last year building context for AI analysts in the mid-market. I’ve open sourced a library from my learnings, along the same principles of the Matt Pocock skills. Principles: * Claude / Codex / Cursor etc are great data agents - you just need context to enable them to answer business…6 points · 4 repliesEdy | Indie BuilderToday's GitHub trending: everyone is building skill packs for AI agents. - mattpocock/skills: skills from Matt Pocock's own .agents folder (+883 today) - obra/superpowers: agent skills framework plus a dev method (+665) - Orkas-AI/Orkas: local-first desktop AI team (+1,044) #buildinpublic #AI6 likes · 1 reply@zyonbuilds@pingdotgg 2. DSH Go puts the DeepSeek Harness agent on your Android phone. → runs on the phone itself → or remote-controls the agent on your computer +11.0k stars this week, 15.0k total. AGPL-3.0. https://t.co/xQbpMcgjSEramozShow HN: WYSIWYG markdown editor for any GitHub reporeplace any github.com url with dunkdown.com https://dunkdown.com/mattpocock/skills4 points · 2 repliesGitHub Trending📈 Today's GitHub Trending: 1. https://github.com/vercel-labs/agent-browser - Rust, 180 stars today 2. https://github.com/cloudflare/security-audit-skill - JavaScript, 2088 stars today 3. https://github.com/google/skills - Python, 449 stars today 4. https://github.com/mattpocock/skills - Shell […]3 likes · 1 reply@iRehmanAhmadmy favourite bits: → /grill-with-docs builds a shared glossary, so the agent stops guessing your jargon → /diagnosing-bugs runs a disciplined loop until the bug is dead → /improve-codebase-architecture finds deep-module candidates → /ask-matt routes you to the right skillremingtonvShow HN: Open-source skill to run Matt Pocock's skills in bulk AFK1 pointMatt Pocock {bot}15 stars until we become the most-starred skills repo in the world... ...and the 13th most starred repo of all time. https://github.com/mattpocock/skills2 likes · 1 reply@LoveAIbrain316,958 star，GitHub 涨速最快的仓库之一，不是框架，是一堆 Markdown。 Matt Pocock 把自己私藏的 .agents 目录开源了：code review、重构、测试，全是生产级 agent skills。 直接复制进 Claude Code 就能用。 https://t.co/iColqIqSZ9 #中文区narwhsterShow HN: A notebook for prototyping with your agentAbove is a link to the NAWC documentation, that's also a live version of NAWC you can try in your browser. I made nawc because I loved spec-driven development with agents, but hated how inconsistent markdown is at describing intent. I needed a more streamlined approach, where intent is captured…1 pointデイリーGitHubトレンド今日のGitHubトレンド mattpocock/skills 実務的なエンジニアリングを実現するためのAIエージェント向けスキル集です。 開発者がコントロールを維持できるよう、小規模で適応性が高く、組み合わせ可能な設計となっています。 特定のモデルに依存せず、Claude CodeやGitHub Copilot、Gemini CLIなどで利用でき、効率的なアプリケーション開発を支援することを目的としています。1 like · 1 reply@abhishebj0github上31.7 万颗星。 这是 Matt Pocock 公开的自己每天在用的 Agent Skills，他说是给认真写工程的人用的，不是 vibe coding。每个都很小，能拆开改成自己的。 Claude Code 一条命令就能装，Codex、Copilot 也有装法。免费开源。 https://t.co/3mc1D2jv25
More from Mindket
See all from MindketPixel Art XL
nerijs · Kit
A pixel art style add-on for Stable Diffusion XL that turns text prompts into pixel-style images, with an optional LCM LoRA for faster runs. For game makers, d…
New
VibeVoice-1.5B
microsoft · Kit
Turns written conversations into spoken audio with up to four speakers and about 90 minutes of speech, an open Microsoft text-to-speech model for research on p…
New
Similar in Developer
See all in Developervoicetut-egyptian-tashkeel
masterofaudio2077 · Kit
Reads Egyptian Arabic text aloud and follows the short vowel marks you write into it. A fine-tune of VoiceTut-TTS that keeps its 19 built-in voices and voice c…
New
Meta-Llama-3.1-8B-Instruct-AWQ-INT4
hugging-quants · Kit
Runs Meta's Llama 3.1 8B Instruct chat model in 4-bit AWQ form, so it fits on a GPU with about 4 GiB of free memory for the weights. For developers who want a…
New
Ru-Jev
besktech · Kit
Scores a Russian-first decision from a context and a list of options, giving a probability for yes/no, choice or rating questions. Runs on a plain CPU, for tea…
New