EasyResearch is an open-source AI Research Team designed to automate the journey from ideas to validated scientific knowledge — enabling researchers to discover, reproduce, improve, and create scientific breakthroughs with autonomous agents.
Skill Claude CodeCodex
Retrieve and verify arXiv paper metadata by arXiv ID, generate BibTeX, inspect abstract/PDF pages with webfetch, and query Semantic Scholar citation/reference metadata. Use as a complement to paper-search after candidate papers are already found; do not use for broad paper search unless paper-search is unavailable.…
Skill Claude CodeCodex
Use proactively when the user asks for autoresearch, auto research, autonomous experiments, overnight optimization, unattended model search, repeated metric improvement, or to keep trying until stopped. Also use when a paper project needs a bounded automatic exploratory campaign before formal validation.
Skill Claude CodeCodex
Use when editing or troubleshooting EasyResearch's own agents, skills, settings.json, models.json, auth.json, extensions, prompts, themes, or .easyresearch configuration; also for API timeouts, retries, transport, compaction, thinking budgets, image settings, mounting skills, or explaining config scope and precedence.…
Skill Claude CodeCodex needs its repo
Publication-figure overlay for draw.io. Use instead of drawio whenever the diagram is for a paper, thesis, dissertation, journal, conference, IEEE/ACM submission, manuscript, camera-ready, Word/LaTeX figure, or other publication. Applies venue, figure-type, color, caption/legend, formula, and paper-readability gates…
Skill Claude CodeCodex
Create, edit, replicate, import, and export draw.io diagrams with an offline YAML-first workflow: architecture, network topologies, flowcharts, UML/ER, org charts, Mermaid/CSV conversion, existing .drawio bundles, style presets, themes, and non-publication formula diagrams. For publication figures (paper, thesis…
Skill Claude CodeCodex
Use when Experiment must build, run, compare, or record reproducible ML/AI paper experiments locally or through the Research Assistant-configured ssh-bash tool and verified SSHFS workspace, including baselines, datasets, controlled trials, multi-seed formal evidence, ablations, and results promotion. Examples: user…
Skill Claude CodeCodex
Use when Experiment must choose experimental units, randomization, blocking, replication, factorial or response-surface structure, controls, or a reproducible allocation before collecting formal evidence.
Skill Claude CodeCodex
Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.
Skill Claude Code
A relentless interview to sharpen a plan or design.
Skill Claude CodeCodex
Use when Experiment must inspect a public Hugging Face dataset's configurations, splits, rows, filters, Parquet shards, size, statistics, Croissant metadata, license, or separately resolved revision before selecting it for evidence.
Skill Claude CodeCodex
Use when Experiment must turn observations or literature limitations into evidence-bounded rival hypotheses, discriminating predictions, controls, measurements, or a preregistration-ready empirical plan before implementation or trials.
Skill Claude CodeCodex
Build LaTeX manuscripts into PDFs, diagnose TeX toolchain errors, choose pdflatex/xelatex/lualatex/latexmk commands, clean build artifacts, and summarize compilation warnings. Use proactively when a user asks to compile, rebuild, export, or troubleshoot a .tex paper, IEEE/ACM/Elsevier template, bibliography, missing…
Skill Claude CodeCodex
Use when Search or Review must resolve a known scholarly identifier, verify metadata or citations, inspect a bounded public scholarly API result, or locate a lawful open-access source beyond broad arXiv/OpenReview discovery.
Skill Claude CodeCodex
Use when Search must turn a selected paper set into a durable, human-readable material handoff for survey writing, experiment design, related-work analysis, or citation checking, especially when refpapers/source.json and converted full text already exist but downstream agents need verified per-paper facts and evidence…
Skill Claude CodeCodex
Search and organize OpenReview and arXiv paper candidates, returning structured results for a user-given topic, time range, and data sources. Use proactively when the user asks to find papers, restrict a time range, prefer OpenReview/arXiv results, or wants examples of what a research direction looks like in target…
Skill Claude CodeCodex
Convert research PDFs and other documents to Markdown using MarkItDown from the EasyResearch skill venv ($EASYRESEARCHVENV). Use proactively when papers must be converted into readable text for agent analysis, especially inside refpapers/pdf to refpapers/text workflows. Examples: user: "Convert these papers to…
Skill Claude CodeCodex
Use when Review must independently assess accepted Markdown or TeX manuscript sources against literature and experiment evidence and produce an immutable source-based review report without editing the submission.
Skill Claude Code
Automate browser interactions, test web pages and work with Playwright tests.
Skill Claude CodeCodex
Use when an empirical paper task will run on one SSH server and the Research Assistant must prepare and verify its remote workspace before dispatching Experiment.
Skill Claude CodeCodex
Use when Writing must assess, draft, revise, or audit an empirical ML/AI method paper whose claims depend on experiment-record.md and results/, including baseline fairness, multi-seed reporting, ablations, formula-rich methods, result tables, citation verification, and derived submission artifacts. Do not use its…
Skill Claude CodeCodex
Use when the Research Assistant must classify and orchestrate a multi-stage paper request, continue an existing project from exact-cwd artifacts, choose between survey, empirical, or hybrid routes, acceptance-review specialist handoffs, or decide whether missing authority, access, evidence, cost, or safety requires a…
Skill Claude CodeCodex
Use when Figures must create or audit a data-driven scientific chart, uncertainty or missing-data display, multi-panel plot, accessible palette, provenance manifest, or publication export rather than a Draw.io diagram.
Skill Claude CodeCodex
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
Skill Claude CodeCodex
Use before a bundled specialist writes a long report or manuscript, recovers an interrupted write, or finishes a normal run or continuation with a durable handoff for its immediate caller.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: