Kipt

Ask Matt skill: find the right Matt Pocock skill for your situation

The ask-matt skill is a router over mattpocock/skills: describe your situation and it names the next skill or flow to run. What it does and how to install it.

By the Kipt team · Updated 24 October 2026

Source: mattpocock/skills · Maintainer: community: Matt Pocock · License: MIT · Checked: 24 October 2026

What is the ask-matt skill?

The ask-matt skill is a router over Matt Pocock's skill collection: you describe the situation you are in, and it tells you which skill or flow to run next and where your decisions sit. It is for people who have installed mattpocock/skills and don't remember what each one is for.

The skill's description reads: "Ask which skill or flow fits your situation. A router over the skills in this repo." Its first line is blunter: "You don't remember every skill, so ask."

The repo's docs for ask-matt answer a request people keep making in the README: a list of the skills in the right order. "This skill is that list." As of 24 October 2026, its skills.sh page listed about 767,100 installs (snapshot, 24 October 2026).

What does it do?

It recommends and stops. Per the docs, it does not grill you, write a spec, open a file or fire the skill it named. What you get back is the next thing to type.

The SKILL.md is a map of flows, paths through the skills, rather than a keyword lookup:

  • The main flow, idea to ship. /grill-with-docs sharpens the idea (or /grill-me with no working directory). A question that needs runnable code detours through /prototype, bridged by /handoff. A multi-session build goes /to-spec, then /to-tickets, then /implement per ticket or /implement-spec for the whole spec. A small build goes straight to /implement. Code is built through /tdd, checked by /code-review, and /retro closes the loop.
  • On-ramps. Incoming bug reports go to /triage, a hard bug goes to /diagnosing-bugs, and an effort too big and foggy for one session goes to /wayfinder.
  • Codebase health. /improve-codebase-architecture finds deepening opportunities, which re-enter the main flow as ideas.
  • Standalones such as /research, /to-questionnaire, /wizard, /wait-what and /teach.
  • A vocabulary layer: /domain-modeling and /codebase-design.

It also gives rules of thumb. Keep grilling, spec and tickets in one unbroken context window, then start each /implement fresh. Don't triage tickets that /to-tickets produced. When a wayfinder map clears, go to /to-spec, not straight to /implement.

The folder has a second file, PHASE-BOUNDARIES.md, covered below, plus agents/openai.yaml (Codex display name "Ask Matt", implicit invocation off).

How do you install it?

The README gives a plugin route and an editable-files route, and says to pick one per agent so you don't get every skill twice. For a router, the plugin route makes the most sense, because a recommendation is only useful if the skill it names is installed.

Claude Code plugin (whole set, updates itself). The collection is listed in Anthropic's official plugin marketplace:

claude plugin install mattpocock-skills@claude-plugins-official

Plugin skills are namespaced, so it appears as /mattpocock-skills:ask-matt.

Codex plugin. From the README:

codex plugin marketplace add mattpocock/skills
codex plugin add mattpocock-skills@mattpocock

Editable files with the skills CLI. The command on skills.sh:

npx skills add https://github.com/mattpocock/skills --skill ask-matt

The skills CLI prompts you to pick skills if you leave out --skill, and -g installs for all projects. Codex and Cursor read the same SKILL.md format; see Codex skills and Cursor skills.

The docs add that the tracker routes (triage, to-spec, to-tickets, implement) assume /setup-matt-pocock-skills has configured an issue tracker, and the router recommends them even before that has happened.

How do you use it?

Type /ask-matt and describe your situation in a sentence or two. The skill sets disable-model-invocation: true, which per the Claude Code skills docs means "Claude can't invoke the skill on its own."

Example prompts (suggestions, not quotes from the repo):

  • /ask-matt I have an idea for a CSV export feature in this repo. Where do I start?
  • /ask-matt Bug reports are piling up from users. What flow do I use?
  • /ask-matt Should I use grill-with-docs or wayfinder for a rewrite of our billing system?
  • /ask-matt We just finished grilling and the context is at 120k tokens. Compact, clear or hand off?
  • /ask-matt just give me the sequence for a multi-session feature build.

The docs say a good answer names what to type and stops, mentions where to clear or compact and where you review, and explains why the nearby alternative is wrong for you. If you already know which skill you want, invoke that skill directly.

What is the phase-boundary tree?

A phase is a chunk of work inside a session, such as grilling, implementation or QA. PHASE-BOUNDARIES.md says the boundary between two phases "is the only place this decision belongs" and gives five options in order, where the first yes wins:

Order Option Take it when
1 Continue The next phase needs this one as a primary source, or there is room left
2 /clear Everything in the session is disposable
3 /handoff Something has to travel: new harness, new directory, a colleague, or a side task
4 Subagent The task can run with you away from the keyboard
5 /compact None of the above

It describes /compact as "the default, not the first reach," and warns that starting there can leave "a fresh session that is confidently wrong about a decision the summary flattened." The file puts the useful window at about 150k tokens on current models. For the Claude Code side of this, see our context window guide and the handoff skill.

Tips and limits

From the repo's docs, not from testing:

  • "Not installed" is usually wrong. User-invoked skills are left out of the skill list the agent sees, so it may report them missing. The docs say sixteen of the plugin's twenty-seven skills carry the flag. Type the command anyway, or check .claude-plugin/plugin.json.
  • It can misdescribe other skills. It may answer from its own one-line summaries. The docs cite a session where it advised skipping to-spec without opening that skill. Ask it to read the target SKILL.md first, and treat the SKILL.md as right when they disagree.
  • It only knows this repo. Proposals for a router over your local skills folder exist, but ask-matt is not that.
  • Edits to installed skills don't last. npx skills update overwrites files and the plugin install is read-only. Put standing rules in your own CLAUDE.md or AGENTS.md.
  • Renames confuse it. Check the changelog before assuming a skill is gone. The docs list to-prd becoming to-spec and writing-great-skills becoming writing-for-agents. They also say pathfinder became wayfinder, but the changelog and the wayfinder docs record the old name as decision-mapping.

As of 24 October 2026, skills.sh showed Pass from Gen Agent Trust Hub, Socket and Snyk. The skill runs no scripts and writes no files. It is MIT licensed.

Save it in your Kipt library

In Kipt, open Skills and paste the mattpocock/skills repo link into the import box. Because the repo holds many skills, Kipt shows a list to pick from, so you can take ask-matt together with the skills it routes to. Your copies live in your own private GitHub repo and install into Claude Code or Codex with one command:

npx skills add https://github.com/<you>/<repo>.git --all -g

Kipt is not affiliated with Matt Pocock or mattpocock/skills.

Similar skills

Questions

What is the ask-matt skill?

ask-matt is an Agent Skill in Matt Pocock's mattpocock/skills repo. It is a router: you describe where you are, and it names the skill or sequence of skills that fits, plus where your decisions sit. It recommends and stops; it does not run the skill it names.

Does ask-matt know about skills from other authors?

No. The repo's docs say it is a hand-written map of the skills in mattpocock/skills, not a scan of what you have installed. It will not route you over your own skills or another author's, and it can lag behind new skills in its own repo.

Why does ask-matt say half the skills aren't installed?

The docs list this as a known, unfixed bug. Most of the skills it routes to set disable-model-invocation: true, so the harness leaves them out of the skill list the agent sees, and the agent reads that list as complete. They are installed: type the slash command anyway.

How do I install the ask-matt skill?

Install the whole collection in Claude Code with claude plugin install mattpocock-skills@claude-plugins-official, or copy just this skill with npx skills add https://github.com/mattpocock/skills --skill ask-matt. The router is only useful if the skills it recommends are installed too.

What does ask-matt say about /compact, /clear and /handoff?

It carries an ordered tree for the boundary between two phases of work: continue if you can, /clear if the context is disposable, /handoff only if something has to travel, a subagent if the task can run unattended, and /compact otherwise. /compact is the default but sits last in the order.

More in Skill directory

All guides

Keep your prompts and skills in one place, in your own GitHub. Install every skill with one command.

Get started

Kipt is an independent product, not affiliated with or endorsed by Anthropic, OpenAI, Cursor or GitHub. Claude, ChatGPT, Codex, Cursor and GitHub are trademarks of their owners.