ツール

Lets agents manage Google Cloud resources in natural language through the gcloud CLI, blocking interactive and other agent-unfriendly commands by default, with configurable allow and deny lists. A preview release from Google, not an officially supported product; install gcloud and sign in first.

Translates articles between Chinese and English or Chinese and Japanese (not English-Japanese) in four steps: analysis, draft, review and final, with options for target audience, style and glossary, aiming for natural-sounding output. No account or API key; URLs are fetched via r.jina.ai.

AWS Labs' official plugin marketplace: deploy-on-aws (architecture recommendations, cost estimates and IaC deployment), aws-serverless, aws-amplify, databases-on-aws, sagemaker-ai and more. AWS suggests Agent Toolkit for AWS for new projects.

165 teaching skills grounded in learning-science research, covering course design, lesson planning, assessment, classroom questioning and student self-study coaching, for teachers and instructional designers. Install all at once as a Claude Code or Codex plugin, or copy them one by one.

Plotting skills for academic papers: bar, line, scatter, radar and other styles distilled from 9 real paper figures. Plot your own data in a chosen style as 300 dpi figures, or reproduce a matplotlib script from a paper screenshot. Needs local Python and matplotlib.

HashiCorp's official plugin marketplace: the terraform plugin has 16 skills (configuration, modules, provider development, testing, import, Stacks and policy), and the packer plugin has 4 image-build and HCP Packer skills.

HashiCorp's official Terraform and Packer skills: write HCL in the official style, refactor modules, terraform test, provider development, Stacks and policy, plus AWS and Azure image builds.

Answers macro questions with market data: combines 13 free sources including Polymarket and Kalshi prediction markets, US Treasury yields, CFTC positioning, SEC insider trades and A-share fund flows into probability estimates, reasoning and scenarios. No API key; needs uv. Not investment advice.

100+ tools that let agents drive COMSOL Multiphysics: build models, set geometry and physics (heat transfer, fluids, electrostatics, solid mechanics), mesh, solve and export results; the accompanying paper is in Neurocomputing. Needs licensed COMSOL 5.x/6.x, Python and Java; install from the repo.

Runs SQL queries against ClickHouse and browses databases and table schemas, read-only by default; can also enable embedded chDB to query data directly in files and URLs. Maintained by ClickHouse.

Jina AI's official remote MCP: converts web pages and PDFs to Markdown, searches the web, arXiv, SSRN and images, and offers screenshots, reranking and semantic dedup. Reading tools work without a key (rate-limited); search and rerank need a Jina API key.

Lets agents browse subreddits, search posts, read comments and analyze user activity, useful for customer sentiment, competitor feedback and market research. Browsing needs no account; search and comments need free Reddit app credentials. Requires Node.js.

Describe what you need and get a typeset PDF: turn handwritten notes or scanned textbooks into LaTeX handouts and cheat sheets, make exams with answer keys, write papers, theses, posters, resumes and letters. 27 templates plus chart and mail-merge scripts. Needs TeX Live (install script included).

Unbabel's machine-translation quality metric: comet-score rates translations per sentence and overall, XCOMET marks error spans and severity, and CometKiwi works without a reference. Good for spot-checking MT. Needs Python 3.8+; models download from Hugging Face, some need a license; GPU helps.

The Docling project's official MCP: converts PDF, Word, PowerPoint and more into structured documents with layout and table detection and OCR for scans, and lets agents build new documents paragraph by paragraph and export Markdown. Local mode needs Python and uv, or use your own Docling Serve.

Rewrites a paper for a target journal: give it 5-8 papers from that journal (plus optional field or lab examples), it builds an editable style guide for your manuscript, then rewrites section by section to the journal's conventions with a change log. PDF input needs a converter such as MinerU.

A research plugin for Claude Code: commands like /omp:survey, /omp:experiment and /omp:write chain literature review, topic selection, experiments, writing and self-review, with 5 specialized agents and 34 research skills that remember project progress across sessions. Experiments lean toward ML.

dbt Labs' official skills: write and edit dbt models, add unit tests and docs, set up the MetricFlow semantic layer, debug failed jobs, and handle dbt version upgrades and cross-platform migrations. Most skills need dbt installed and an existing project.

dbt Labs' official plugin marketplace with dbt and dbt-migration plugins: write models and unit tests, build semantic-layer metrics, govern with dbt Mesh, debug failed jobs, and handle version upgrades and cross-platform migrations. Most skills need an existing dbt project.

21 English writing rules (12 from The Elements of Style, Orwell, Pinker and other classics, 9 from observed LLM habits) for clearer, leaner papers, grant proposals, design docs and documentation. Expository English only; not for fiction or marketing copy.

Two skills for before and after peer review: review-revision checks a manuscript as a senior reviewer would and plans section-by-section revisions; rebuttal-response triages reviewer comments, plans extra evidence and drafts point-by-point replies. Takes PDF, LaTeX, Word or Markdown.

The five-agent PaperOrchestra pipeline (outline, figures, literature review, section writing, polishing) as skills: turns research ideas, experiment logs and a conference template into a LaTeX paper and checks citations via Semantic Scholar. Needs Python; verify the draft yourself.

Economics writing skill built on 50+ guides (Cochrane, McCloskey and others): shapes abstracts, introductions and conclusions by paper type and 13 identification strategies such as DiD, IV and RDD, simulates three referee scores and flags citations to verify.

Helps agents write geospatial SQL for PostGIS, DuckDB, BigQuery and Snowflake: inspect schemas and keep query costs in check, then render results as interactive maps to review and correct. Needs Python and your own database or warehouse; maps use Dekart, which can run locally in Docker.