research
Skill2 starsChanged 12 days agoMITView on GitHub
Research how to build or integrate something before the work starts: how it works, the options, the APIs to use, and the risks, from primary sources, as a cited report in the repo. Use when the user runs /research, asks to research, look up or compare something, asks how to connect or integrate a service, or another skill needs facts from docs, specs or the web.
189 other projects use a version of this.
Adopted by
Most recently taken up by
- Suge8/friedbuntheir own version, first seen Nov 15, 2026
- Fractera/Agent-Engineering-Infrastructuretheir own version, first seen Nov 15, 2026
- Fractera/Agentic-Engineering-Infrastructuretheir own version, first seen Nov 15, 2026
- iha2/superlitetheir own version, first seen Nov 15, 2026
- SireJeff/k0ntexttheir own version, first seen Nov 15, 2026
- quickchatai/quickreviewtheir own version, first seen Nov 15, 2026
Used with
Projects with it also use daily-arxiv, exp-eval and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
LambenthanSkilldaily-arxiv
每日从 arXiv 拉取新论文,过滤相关性,auto-ingest 高优先级论文,检测 SOTA 更新
Also in 5 projects that use this
LambenthanSkillexp-eval
实验判决门:Review LLM 独立评判实验结果 → 4 种判决路径 → 自动更新 claims confidence、ideas status、graph edges
Also in 5 projects that use this
truongtv22SkillProblem-Solving Techniques
Apply systematic problem-solving techniques for complexity spirals (simplification cascades), innovation blocks (collision-zone thinking), recurring patterns (meta-pattern recognition), assumption constraints (inversion exercise), scale uncertainty (scale game), and dispatch when stuck. Techniques derived from Microsoft Amplifier project patterns adapted for immediate application.
Also in 5 projects that use this
liufeng1976Skillsetup-ts-deep-modules
Wire dependency-cruiser into a TypeScript repo so each package is a deep module — implementation hidden in subfolders, reachable only through its entry-point files. User-invoked.
Also in 43 projects that use this
skyllwtSkillexp-pilot-eval
Pilot result evaluation — read pilot results, apply success criteria, update idea page (pilot_result, failure_reason if failed), generate PILOT_VERDICT_REPORT. Called by /ideate Phase 5 after /exp-pilot-run.
Also in 4 projects that use this
skyllwtSkillexp-pilot-run
Pilot experiment execution — read Pilot Spec YAML, write pilot code, run experiment(Confirm with the user before operation and require the applicant to conduct manual inspection), return results. Called by /ideate Phase 5. Does NOT modify wiki pages or judge pass/fail.
Also in 4 projects that use this
- +5 more from Lambenthan
- +1 more from skyllwt
Other versions189 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 189
- mattpocock/skills★ 325ksame name
- Yeachan-Heo/oh-my-claudecode★ 41ksame name
- Yeachan-Heo/oh-my-claude-sisyphus★ 41ksame name
- QuivrHQ/quivr★ 40ksame name
- The-Vibe-Company/quivr★ 40ksame name
- udecode/plate★ 17ksame name
- openai/plugins★ 7.8ksame name
- ArcReel/ArcReel★ 6.6ksame name
- vinvcn/mattpocock-skills-zh-CN★ 5.8ksame name
- withcoral/coral★ 4.9ksame name
179 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository22 other files it gives its agents.
The file
We split this skill into the rules it gives an agent
7 rules. 0 are common, 5 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 170 edited versions we read keep these.
5
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 27
Ask with the question tool only when the answer changes what the researchers look for: a term with two meanings, a missing version, country …
- Distinctive: line 42
When another skill calls this one, take its question and constraints as given, and ask nothing.
- Very specific: line 52
When a report on the same topic exists, keep it.
- Very specific: line 60
Start each researcher with the Agent tool, subagent_type: general-purpose.
- Distinctive: line 89
Read the summary of each researcher.
- Distinctive: line 119
Read the report. Leave it uncommitted: the user decides whether it belongs in the repo history.
- Distinctive: line 121
When another skill called this one, give back the report path and the answer: the answer paragraph of a survey or an integration, or the …
“Common” is measured: kept by at least 6 in 10 of the 170 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Research
/research answers a question with a cited report that the user can build from. You are the coordinator: you scope the question, start the researchers, and give their notes to a writer. The subagents search and read. You keep your context for the user's work.
Each claim in the report traces to a primary source: the docs, spec, regulation, source code, changelog or first-party API that owns the fact. A blog post, forum answer or AI summary that repeats the fact is a lead to the owner, not a citation.
1. Scope
Read the question. Then read the stack of the app, when the work runs in a repo: the language, the framework, the hosting, and each integration that is already there, such as a payment or email provider. Take it from the manifest files, such as package.json, and the config. The stack is a constraint for each researcher: an SDK for the wrong language is no answer.
Ask with the question tool only when the answer changes what the researchers look for: a term with two meanings, a missing version, country or date, or a name that points to two things. Ask two questions at most.
Then pick the kind:
- Integration: how to connect the app to a service, an authority or a standard, such as SAT e-invoicing, a bank feed or an OAuth provider. A survey with the four facets below.
- Lookup: one set of facts with one owner, such as how an API pages its results or what a config flag does. One researcher, and its notes are the report.
- Survey: a comparison or a question with distinct facets that is not an integration. One researcher per facet, 3 to 5. The facets do not overlap, and together they cover the question.
The facets of an integration:
- Process: how the integration works end to end. The parties, the flow of data, the rules or regulation that apply, and the credentials, certificates or registrations that the app or the user must get.
- Options: each way to integrate. Direct with the owner, through each main provider, or with an open source library. For each option: cost, limits, support, and the fit with the stack.
- APIs: for each option that fits, the endpoints or SDK, the auth, the data formats, the rate limits, the sandbox, and the library for the stack.
- Operation: testing, errors and retries, failure modes, compliance duties such as retention, and each deprecation or deadline that is announced.
When another skill calls this one, take its question and constraints as given, and ask nothing.
Completion: you have the question in one sentence, the stack, the constraints and the kind. For a survey, you have each facet with its key questions.
2. Place
Pick a slug of 2 to 5 lowercase words, joined by hyphens, such as sat-cfdi-integration.
The report goes where the repo keeps such notes. Look for a folder such as docs/research/ or docs/notes/, and match its convention. With none, use docs/research/<slug>.md. Outside a repo, use research/<slug>.md in the working directory.
When a report on the same topic exists, keep it. Give the new report a slug that says what this run adds, or <slug>-2. The earlier report is background for the writer.
The notes go in <scratchpad>/research/<slug>/, because they are working files, not repo content.
Completion: you have the absolute report path, which names no existing file, and the absolute notes folder. The parent folders of both exist.
3. Research
Start each researcher with the Agent tool, subagent_type: general-purpose. Start all of them in one message, so that they run in parallel. Give each one this prompt:
Research <narrow topic>.
Objective: <what the notes must cover>
Key questions:
- <question>
Where to look: <the primary sources that own these facts, such as an official portal, a docs site, a regulation, a repo, release notes, or files in this repo>
Stack: <language, framework, hosting, existing integrations, or "none">
Constraints: <version, country, date and other bounds from the question>
Save your notes to <absolute path>
First, read <absolute path of this skill's folder>/RESEARCHER.md and follow it.For a lookup, the save path is the report path. For a survey, it is <notes folder>/<facet>.md.
Tell the user in one line what runs, then go on with the user's other work until each researcher reports back.
Completion: each researcher reported back, and each file that you named exists.
4. Gaps
Read the summary of each researcher. A blocking gap is a key question with no cited answer, where the report cannot answer the user's question without it. For blocking gaps, start one more round of researchers, on the gaps only.
One gap round at most. After it, go to step 5 with what you have. A gap that is still open goes into the report as a gap.
Completion: no gap blocks the answer, or the gap round reported back.
5. Write
For a lookup, go to step 6.
For an integration or a survey, start one writer with the Agent tool, subagent_type: general-purpose:
Read the notes in <notes folder> and write a report that answers: <the question in full>.
Kind: <integration | survey>
Stack: <the stack from step 1, or "none">
Earlier report to build on: <absolute path, or "none">
Save the report to <report path>.
First, read <absolute path of this skill's folder>/WRITER.md and follow it.Completion: the report exists at the report path.
6. Deliver
Read the report. Leave it uncommitted: the user decides whether it belongs in the repo history.
When another skill called this one, give back the report path and the answer: the answer paragraph of a survey or an integration, or the takeaways of a lookup. That skill goes on from there.
Report
Report path · kind · the answer in 3 to 5 sentences · each open gap · next step: for an integration, /issue idea or /spec, with the report path as input.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

