Executable knowledge

Don't start every research task from zero.

PaperTrace Skills package repeatable research practice into inspectable workflows. Use the built-in publishing skills, or connect a larger skill library when the task needs specialist execution.

Runnable here

Research and publishing workflows

4 skills
01SKILL.md

Paper → Deep Dive

Turn a research paper into a bilingual explanation with formula walkthroughs, concrete examples, connections and an interactive demo.

Understand → explain → visualize → verify

Installed · .agents/skills/papertrace-deep-dive

02SKILL.md

Daily Research Signal

Collect recent work, identify why it matters, label one editor's pick and publish a compact bilingual digest.

Collect → rank → summarize → source-check

Installed · .agents/skills/research-signal-curator

03SKILL.md

Weekly Narrative

Find the week's connecting idea instead of listing links, then group 5–10 papers around one emerging narrative.

Collect → cluster → theme → publish

Installed · .agents/skills/weekly-research-narrative

04EXTERNAL

AI Research Loop

An installed MIT-licensed workflow for hypotheses, literature review, baseline reproduction, leak-free experiments, multi-seed analysis and evidence-backed writing.

Frame → reproduce → run → analyze → write

Toadoum/ai-research-skill · Installed · source + scripts inspected

Open ecosystem

Reuse before rebuilding

A collection is admitted only when its source, license, install path and execution boundary are visible. Large catalogs stay behind a router instead of flooding every prompt.

Catalog checkedApache-2.0 collection

AREX Research Skills Library

1,060 public operating skills · 170 repositories · 14 families

A routed repository-skill graph for ML, data, evaluation, retrieval, vision, generation, MLOps and scientific computing. The router loads only the relevant fragment.

Inspect source ↗
Installed + inspectedMIT

AI Research Loop

Hypothesis → experiment → paper

Installed project-locally after inspecting the full manifest and both deterministic Python helpers. It writes only experiment scaffolds and an explicit lessons log.

Inspect source ↗
Source checked · selective installMIT

AI4S Skills

7 connected skills · exploration to integrity audit

A source-checked suite for topic exploration, literature surveys, experiment packages, paper writing, mind maps and integrity audits. Listed for selective review; not installed wholesale.

Inspect source ↗
Catalog checked · per-skill reviewPer-skill license

Scientific Agent Skills

156 skills · databases, packages and scientific workflows

Broad cross-domain coverage with unusually explicit boundaries for evidence, ethics, clinical use and local deterministic helpers. Individual skills still require inspection before installation.

Inspect source ↗
Discovery directoryMixed upstream licenses

Awesome Scientific Skills

Curated directory · science, medicine, finance and data

A discovery index pointing to skills from many upstream projects. It is useful for breadth, but each linked skill keeps its own license and trust boundary.

Inspect source ↗
Skill architecture

Load expertise only when it is needed

A small routing layer chooses a scoped, source-backed workflow. The workflow produces an artifact, then verification gates decide whether it can be published.

Goal→Route→Skill→Verify→Publish