Source: mattpocock/skills · Maintainer: community: Matt Pocock · License: MIT · Checked: 24 October 2026
What is the research skill?
The research skill hands reading work to a background agent: it investigates a question against primary sources and leaves one cited Markdown file in your repo, while your own session keeps going. It is for developers who need an outside fact, such as how a third-party API behaves, before they can make a decision.
It comes from mattpocock/skills, which Matt Pocock describes in the README as "My agent skills that I use every day to do real engineering - not vibe coding." The skill's description reads: "Investigate a question against high-trust primary sources and capture the findings as a Markdown file in the repo. Use when the user wants a topic researched, docs or API facts gathered, or reading legwork delegated to a background agent."
As of 24 October 2026, its skills.sh page listed about 711,500 installs (snapshot, 24 October 2026).
What does it do?
The whole SKILL.md is three short rules. It opens with "Spin up a background agent to do the research, so you keep working while it reads." That agent's job:
- Investigate the question against primary sources ("official docs, source code, specs, first-party APIs"), not a secondary write-up, and follow every claim back to the source that owns it.
- Write the findings to a single Markdown file, citing each claim's source.
- Save it where the repo already keeps such notes. If there is no convention, put it somewhere sensible and say where.
The folder holds SKILL.md and agents/openai.yaml, which gives Codex a display name ("Research") and a short description.
The research docs describe it as "legwork you delegate, not thinking you outsource." You still make the decision; the file is what you plan or argue against.
When to reach for it instead of another skill
The docs give a routing table. A shortened version:
| What you need | Skill |
|---|---|
| An outside fact a decision is waiting on | research |
| A decision made with you, by interview | grilling |
A lasting architecture decision in GLOSSARY.md and ADRs |
grill-with-docs |
| To find out whether an approach works in your codebase | prototype |
| A plan too big for one session | wayfinder |
Wayfinder is the one skill that calls research directly. The research docs say a wayfinder charting session spawns one /research subagent per research ticket and runs them in parallel, each recording findings on a throwaway research/<name> branch.
How do you install the research skill?
The README offers a plugin route and an editable-files route, and says to pick one per agent "because installing both gives you every skill twice."
Claude Code plugin (whole set, updates itself). From Anthropic's official plugin marketplace:
claude plugin install mattpocock-skills@claude-plugins-official
Plugin skills are namespaced as /plugin-name:skill-name, so it appears as /mattpocock-skills:research.
Codex plugin. From the README:
codex plugin marketplace add mattpocock/skills
codex plugin add mattpocock-skills@mattpocock
Just this skill, as editable files. The command on skills.sh:
npx skills add https://github.com/mattpocock/skills --skill research
Add -g to install for all projects. In Claude Code the folder lands in .claude/skills/research/ in your project or ~/.claude/skills/research/ for personal skills, the standard skill locations. Codex and Cursor read the same SKILL.md format; see Codex skills and Cursor skills.
The research docs list no prerequisite skills, and the skill doesn't use an issue tracker, so the tracker settings from /setup-matt-pocock-skills don't affect it.
How do you use it?
The skill is model-invoked, so the agent may load it when a task turns into reading legwork. You can also type /research. In Claude Code, a background subagent runs while you keep working; the subagents docs cover how background tasks work and how to stop one from /tasks.
The docs' main advice is scope: "A narrow, answerable question (one API, one behaviour, one version claim) comes back far better than 'research X'."
Example prompts (suggestions, not quotes from the repo):
/research Does the Stripe Node SDK retry idempotent requests by default, and how many times?/research What changed in the Next.js caching defaults between our version and the latest? Cite the release notes.Research whether Postgres 17 supports MERGE ... RETURNING. Save it with our other notes in docs/research.Before we pick an auth library, research how Better Auth handles session revocation. Primary sources only.
When it finishes, open the file and follow two or three citations. The docs' test: if a link lands on a summary of the thing rather than the thing, the run failed.
Tips and limits
From the repo's docs, the issue tracker and skills.sh, not from testing:
- It can spawn a second agent. The background agent may read "Spin up a background agent" as addressed to itself. Issue #530 reported this and was closed as not planned on 6 October 2026; the repo's out-of-scope note says "The harness should be responsible for stopping infinite subagent loops, not skills." In Claude Code, the subagents docs say to set
CLAUDE_CODE_MAX_SUBAGENT_SPAWN_DEPTHand "Set1to turn nesting off." Otherwise, watch your task list after invoking it and stop any duplicate. - It can silently do nothing. The docs say that if your own instructions forbid agents from re-delegating, the background agent may decline the task.
- The model decides what counts as high-trust. There is no allowlist or verification pass. Your protection is the citation on each claim.
- No stopping rule. Broad questions lead to runs that go too deep or miss the detail that mattered.
- Files are not reused automatically. Nothing loads a past research file. Feed it into the next step yourself, for example by attaching it to a spec.
- Think before committing the file. The docs report the community view that research notes go stale and "can poison future repo reads," and that ADRs are what belongs in git.
- Security audits. On skills.sh, Gen Agent Trust Hub, Socket and Snyk all rated it Pass as of 24 October 2026.
The skill runs no scripts of its own; the background agent reads sources and writes one Markdown file. Under the MIT License you can copy and change it as long as you keep the copyright notice.
Save it in your Kipt library
Because the SKILL.md is only a few lines, research is an easy skill to adapt, for example by naming the folder your team uses for notes or adding a stop rule. In Kipt, open Skills and paste the skill's GitHub folder link into the import box, or drop the folder or a .zip. For a repo with many skills, Kipt shows a list to pick from. Your skills 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
- Wayfinder skill: charts a large plan and fires research tickets as subagents.
- Grilling skill: turns the facts into a decision by interviewing you.
- Grill with docs skill: records lasting decisions in a glossary and ADRs.
- To-tickets skill: slices a settled plan into tickets once the facts are in.
- Dispatching parallel agents skill: the Superpowers pattern for sending independent work to several agents at once.