research
Skill1 starsChanged 3 months agoMITView on GitHub
Research a topic and save structured findings. Use when the user wants to find papers, search for academic sources, explore a topic, look up literature, find references, or gather evidence. Triggers on "find papers on", "what does the literature say about", "search for sources on", "is there research on".
- Reads credentials
189 other projects use a version of this.
Adopted by
Most recently taken up by
- bamlab/flashlighttheir own version, first seen Nov 16, 2026
- lhotwll217/owner-operatortheir own version, first seen Nov 16, 2026
- 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
Used with
Projects with it also use Problem-Solving Techniques, daily-arxiv 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.
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 6 projects that use this
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
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 49 projects that use this
parcadeiSkillast-grep-find
AST-based code search and refactoring via ast-grep MCP
Also in 4 projects that use this
SireJeffSkillcontext-optimize
Context Optimization - Orchestrate RPI workflow TODO generation with interactive scoping
Also in 4 projects that use this
- +5 more from Lambenthan
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
- The-Vibe-Company/quivr★ 40ksame name
- QuivrHQ/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★ 5ksame 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 repository13 other files it gives its agents.
- .claude/CLAUDE.mdCLAUDE.md
- factsSkill
- figureSkill
- full-reviewSkill
- ingestSkill
- onboardSkill
- plan-sectionSkill
- reviewSkill
- save-materialSkill
- statusSkill
- teachSkill
- verify-sourcesSkill
- writeSkill
The file
We split this skill into the rules it gives an agent
9 rules. 0 are common, 5 are distinctive, and 4 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.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 12
Read .claude/CLAUDE.md and docs/requirements.md first so you know what the document is about and what counts as an acceptable source here.
- Very specific: line 16
Search resources/ and docs/research/ before searching the web.
- Distinctive: line 20
Use several methods. Any one of them alone has a characteristic blind spot.
- Distinctive: line 24
Search the open web and the open indexes for the topic.
- Very specific: line 62
Check .claude/rules/research.md and .env for the databases in scope and their proxied URLs (DB_SCOPUS_URL, DB_WEB_OF_SCIENCE_URL, …
- Distinctive: line 66
Stop. Do not type anything into any credential field, and do not read the page for stored credentials.
- Distinctive: line 68
Wait for them. Do not retry, do not navigate away, do not try another route into the database.
- Distinctive: line 69
When they confirm, take a fresh snapshot and continue from the authenticated page.
- Very specific: line 90
Write to docs/research/research_<topic>_<YYYY-MM-DD>.md in the format set by .claude/rules/research.md.
“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 — Find and evaluate sources
Your job is to come back with a small number of real, relevant, high-quality sources and an honest account of how you found them. A long list padded with plausible-looking references is worse than nothing, because every entry has to be checked by hand later.
Read .claude/CLAUDE.md and docs/requirements.md first so you know what the document is about and what counts as an acceptable source here.
1. Check what you already have
Search resources/ and docs/research/ before searching the web. Prior sessions may already have covered this topic or an adjacent one. Say what you found and build on it rather than starting over.
2. Search
Use several methods. Any one of them alone has a characteristic blind spot.
Method A — Keyword search, for breadth
Search the open web and the open indexes for the topic. Two habits matter more than the choice of engine:
- Phrase-search multi-word concepts.
"learning to rank"finds the field;learning to rankfinds sentences about ranking that happen to contain "learning". - Search titles when you want precision and full text when you want recall. Run both and compare.
Vary the vocabulary deliberately. Fields name the same idea differently, and the paper you need may use the term you did not think of.
Method B — Citation chains through open APIs, for depth
This is the highest-yield method and it should be part of every session. Find one or two strong anchor papers, then walk the citation graph outward in both directions.
OpenAlex — best for bulk search, filtering, and metadata. No key required; an email address in OPENALEX_MAILTO puts you in the faster polite pool.
Search: https://api.openalex.org/works?search={query}&per_page=25&sort=cited_by_count:desc&mailto={email}
By DOI: https://api.openalex.org/works/doi:{doi}
Filtered: https://api.openalex.org/works?filter=default.search:{query},type:article,from_publication_date:2015-01-01&sort=cited_by_count:descEach work carries cited_by_api_url (fetch it for papers citing this one) and referenced_works (OpenAlex IDs of what it cites).
Semantic Scholar — best for the citation graph itself and for abstracts. Works unauthenticated at about one request per second; SEMANTIC_SCHOLAR_API_KEY raises that, sent as the header x-api-key.
Search: https://api.semanticscholar.org/graph/v1/paper/search?query={q}&fields=title,authors,year,citationCount,externalIds,venue,abstract&limit=20
By DOI: https://api.semanticscholar.org/graph/v1/paper/DOI:{doi}?fields=title,authors,year,citationCount,venue,abstract,references,citations
Citing it: https://api.semanticscholar.org/graph/v1/paper/{id}/citations?fields=title,authors,year,citationCount,venue&limit=50
It cites: https://api.semanticscholar.org/graph/v1/paper/{id}/references?fields=title,authors,year,citationCount,venue&limit=50Crossref for authoritative bibliographic metadata, Unpaywall (https://api.unpaywall.org/v2/{doi}?email={email}) to find a legal open-access PDF when the publisher version is locked.
Snowball procedure: anchor paper → fetch its citations and references → sort by citation count → read the titles → pick the two most promising and repeat. Two levels is usually enough. Note that citation count measures uptake, not correctness, and it penalises anything published in the last two years.
Method C — Subscription databases through a browser, for quality
Scopus, Web of Science, EBSCO, JSTOR, and ProQuest index peer-reviewed literature with structured metadata and no preprint noise. They sit behind an institutional login, so no API key reaches them. Drive them with the Playwright MCP browser instead.
Check .claude/rules/research.md and .env for the databases in scope and their proxied URLs (DB_SCOPUS_URL, DB_WEB_OF_SCIENCE_URL, DB_EBSCO_URL, DB_OTHER_URL). If only LIBRARY_PROXY_SUFFIX is set, construct the URL by appending the suffix to the publisher host, e.g. www.scopus.com becomes www-scopus-com.ez.example.edu.
The login flow. Navigate to the search form. Many proxies pass authentication through automatically. If a login page, institution picker, or SSO screen appears instead:
- Stop. Do not type anything into any credential field, and do not read the page for stored credentials.
- Tell the user exactly what is on screen and what to do: sign in themselves in the browser window that is already open, and say when they are through.
- Wait for them. Do not retry, do not navigate away, do not try another route into the database.
- When they confirm, take a fresh snapshot and continue from the authenticated page.
The user's credentials are theirs. You never see them, never store them, and never ask for them in chat.
Searching, once you are in. The mechanics differ per database, but the pattern holds: choose the field (title/abstract/keywords is usually right, title-only for precision), enter phrase-quoted terms, run several term combinations rather than one, sort by citation count to find what the field has accepted, and read the abstract before recording anything. Follow the "cited by" and "references" links for the same snowball procedure as Method B. Note any per-database syntax quirks in .claude/rules/research.md as you learn them, so the next session starts ahead.
3. Record each source
- Authors, year, title, venue
- DOI or a resolvable URL — required
- Key findings, and the method that produced them
- Which section of the document this supports, and what claim it would back
- Quality tier (below), citation count, and whether it is peer-reviewed
- How you found it: which search, or cited by which paper
4. Document the search itself
In the research file, record the databases searched, the exact query strings, roughly how many results you scanned, and the date. This makes the search reproducible in spirit and is often required in the methodology chapter of a literature-heavy document.
5. Save
Write to docs/research/research_<topic>_<YYYY-MM-DD>.md in the format set by .claude/rules/research.md.
6. Recommend next steps
Which sources are worth a full summary via /save-material, which anchors are worth another round of snowballing, what you could not find, and what remains unanswered.
Quality bar
Reject unless all three hold:
- The source exists, confirmed in at least one index or on the publisher's site. Models generate convincing citations to papers that were never written. Verify before you write it down.
- It has a DOI or a resolvable URL. Without one, mark it unverified and do not recommend it.
- Its publication year is not in the future.
Tier every source you keep:
- Tier 1 — peer-reviewed journal or major conference. Prefer these.
- Tier 2 — credible practitioner or preprint: well-cited arXiv work, engineering publications from organisations with real data. Useful, but say plainly that they are not peer-reviewed.
- Tier 3 — workshop papers, minor venues, lightly cited preprints. Use to fill a specific gap Tiers 1 and 2 leave open.
- Tier 4 — no DOI, working papers, repository-only uploads, popular press. Flag, do not build an argument on it.
Be honest about the limits of the search. Ten verified sources with an accurate account of what was not covered beats forty of uncertain provenance.
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.

