gxCaesar/open-research-skills

从选题到可复现发布：十个可独立安装的科研 agent skill —— 顶会/期刊论文、科研绘图、基金 proposal 撰写、方法迭代到 SOTA、冷审与代码数据发布。Ten installable agent skills for research: scientific figures, conference and journal manuscripts, research funding proposals, method development, cold review and artifact release.

10 skills listed · Apache 2.0 · 63 stars · last pushed 23 Sept 2026

gxCaesar/open-research-skills on GitHub

Across these 10 skills

Pattern check

Nothing flagged in 9

Items to review in 1: build-scientific-visualizations

What it can touch

10 bundle scripts
8 run shell commands
6 name web addresses
+2 more
- 5 reach the network
- 1 includes files we can't read

Checked at commit c79bac6 on 3 Nov 2026.

Install the collection

1 of these skills has items to review. Read them before you install the whole collection.

Claude Code plugin

Installs the latest version, not the one we checked. Installs everything in the plugin, including any hooks, MCP servers and scripts our check doesn't cover.

claude plugin marketplace add gxCaesar/open-research-skills --scope user &&claude plugin install open-research-skills@open-research-skills --scope user

Any agent

Installs the latest version, not the one we checked. The skills CLI reports installs to skills.sh unless you set DISABLE_TELEMETRY=1.

npx skills@1.7.4 add gxCaesar/open-research-skills

Skills

All10 Items to review1

Skills in gxCaesar/open-research-skills
Skill	What it can touch	Pattern check	Licence	Updated
build-scientific-visualizationsUse when creating, redesigning, planning, or visually auditing scientific figures, flowcharts, method schematics or model-architecture diagrams, graphical abstracts, dense many-panel compound figures, domain-specific omics visualizations, and conference or journal figure sets, including Nature-style main, Extended Data, and Supplementary figures. Independently owns visual construction, figure legends, source-data preparation, and visual QA; does not rewrite the scientific analysis or manuscript claims. Runs shell commands +4 more Pattern check: 11 items to reviewApache 2.0Updated 23 Sept 2026	Runs shell commands Reaches the network +3 more	Pattern check: 11 items to review	Apache 2.0	23 Sept 2026
develop-method-to-sotaUse when a method is not yet beating its baseline and the next change has to be chosen, including improving a model, reaching state of the art, deciding which component to add next, measuring headroom before building, designing an ablation control that can void a component, reading error by slice instead of one aggregate, or deciding whether to keep iterating or stop. Use during exploratory development on a development split, before any confirmatory run is frozen. Runs shell commands +1 more Pattern check: Nothing flaggedApache 2.0Updated 9 Sept 2026	Runs shell commands Bundles scripts	Pattern check: Nothing flagged	Apache 2.0	9 Sept 2026
prepare-conference-manuscriptsUse when preparing, drafting, revising, polishing, rebutting, auditing, anonymizing, packaging, or checking a conference manuscript, abstract, appendix, supplement, rebuttal, author response, camera-ready paper, or conference statistical or evaluation report for AAAI, ICLR, ACL, CVPR, ICML, NeurIPS, or another peer-reviewed computing venue. Also use for single-section or section-by-section polish and resuming the next section. Includes manuscript-context figures and venue-package checks without requiring another skill; disputed experiment evidence must be resolved before changing manuscript prose. Runs shell commands +3 more Pattern check: Nothing flaggedApache 2.0Updated 15 Sept 2026	Runs shell commands Reaches the network +2 more	Pattern check: Nothing flagged	Apache 2.0	15 Sept 2026
prepare-journal-manuscriptsUse when preparing, drafting, revising, polishing, auditing, or packaging a journal manuscript, abstract, statistical or evaluation report, cover letter, editor-facing submission pack, data availability statement, peer-review response, revise and resubmit package, clean manuscript, marked manuscript, or final accepted-paper files. Also use for single-section or section-by-section polish and resuming the next section. Includes manuscript-context figures and journal-package checks without requiring another skill; disputed experiment evidence must be resolved before changing manuscript prose. Runs shell commands +3 more Pattern check: Nothing flaggedApache 2.0Updated 15 Sept 2026	Runs shell commands Reaches the network +2 more	Pattern check: Nothing flagged	Apache 2.0	15 Sept 2026
release-research-artifactsUse when packaging the code and data a reader or reviewer will download, including releasing the code for a paper, building an anonymized repository for double-blind submission, preparing an archive with a DOI for a journal, writing citation metadata, producing a data card or model card, or preparing a submission for artifact evaluation. Use once the results are frozen, so that what ships is what produced them. Runs shell commands +2 more Pattern check: Nothing flaggedApache 2.0Updated 10 Sept 2026	Runs shell commands Names web addresses +1 more	Pattern check: Nothing flagged	Apache 2.0	10 Sept 2026
research-publication-pipelineUse when selecting, starting, or resuming an evidence-gated research route for a top computer-science venue or Nature-family journal; source-grounded paper reading, AI-for-biology data feasibility, scoop checks, venue fit, baseline headroom, frozen protocols, controlled iterations, claim locking, negative-result handling, preparing a full working manuscript with figures, or curating a reproducible public code release. The workflow completes its local research-to-manuscript scope without another skill; compute, locked-test access, canonical mutation and external writes require their own authorization. Runs shell commands +3 more Pattern check: Nothing flaggedApache 2.0Updated 14 Sept 2026	Runs shell commands Reaches the network +2 more	Pattern check: Nothing flagged	Apache 2.0	14 Sept 2026
review-others-manuscriptsUse when acting as an invited referee on someone else's submission, including writing a referee report, drafting reviewer comments, assessing a revision round against the authors' response letter, running an editorial prescreen for scope and policy, or judging whether a stated claim is carried by the evidence in a manuscript under review. Use for a manuscript that is not yours and is confidential; rehearsing objections against your own unsubmitted paper is a different task. Bundles scripts Pattern check: Nothing flaggedApache 2.0Updated 9 Sept 2026	Bundles scripts	Pattern check: Nothing flagged	Apache 2.0	9 Sept 2026
run-cold-review-panelUse when a manuscript you wrote is nearly finished and needs a hostile read before it is submitted, including a mock review, a simulated reviewer panel, a pre-submission review, red-team the paper, asking whether this would get rejected, or checking which reviewer objection the paper cannot survive. Use for your own unsubmitted manuscript; reviewing someone else's submission as an invited referee is a different task. Reaches the network +1 more Pattern check: Nothing flaggedApache 2.0Updated 10 Sept 2026	Reaches the network Bundles scripts	Pattern check: Nothing flagged	Apache 2.0	10 Sept 2026
survey-and-audit-noveltyUse when deciding whether a research direction is still open, including a literature survey that has to end in a decision, a novelty audit, a prior art check, a scoop check against recent work, asking whether anyone has already done this, or checking that the variables an idea needs actually co-exist in one obtainable dataset. Use before an experimental plan is written, and again whenever a near-identical paper appears. Runs shell commands +1 more Pattern check: Nothing flaggedApache 2.0Updated 10 Sept 2026	Runs shell commands Bundles scripts	Pattern check: Nothing flagged	Apache 2.0	10 Sept 2026
writing-funding-proposalsUse when writing a research funding proposal: establishing current policy or authority, organizing evidence, mapping arguments, planning proposal-specific technical routes and figures, preparing section briefs, reviewing, or validating an application to NSFC or to a provincial fund, including Guangdong. Other funders remain supported through the generic workflow. Includes permitted local figure construction without another skill. Check current funder and institutional authoring policy before producing prose; when direct generation is prohibited or unverified, limit work to source verification, evidence maps, section briefs, structural critique, and audit. Runs shell commands +2 more Pattern check: Nothing flaggedApache 2.0Updated 10 Sept 2026	Runs shell commands Names web addresses +1 more	Pattern check: Nothing flagged	Apache 2.0	10 Sept 2026