Why this matters
Science needs shareable method layers, not just isolated tool use.
Operational knowledge is still too private
Valuable work often depends on tacit prompts, undocumented shell history, and fragile habits that do not survive handoff across people, labs, or time.
Powerful scientific tools remain hard to operationalize
APIs, command-line programs, and data resources are often strong in capability but weak in day-to-day transferability. The missing layer is frequently method packaging rather than new algorithms.
Reusable skills can turn one-off routines into infrastructure
A published skill can preserve exact commands, references, output expectations, and safety constraints while making a workflow easier to invoke and easier to teach.
Platform Structure
Scientific Tooling separates platform identity from project identity.
1
Organization
Scientific Tooling serves as the umbrella brand and public-facing site.
2
Repository
Structured Intelligence is the flagship repository and active implementation surface.
3
Manuscripts
Formal web editions translate reusable workflows into scientific prose.
4
Reuse
Researchers can inspect, rerun, critique, and extend the operational layer.
The point of this structure is clarity. The domain and organization describe the broader initiative; the project path identifies the current flagship implementation.
Current Projects
Three repositories, one publication layer, one umbrella initiative.
Flagship Repository
Structured Intelligence
54 reusable local skills across NGS, RNA-seq, single-cell RNA-seq, metagenomics, statistical analysis, protein thermostability prediction and design, public database search and download, and literature retrieval — orchestrated by two expert agents.
- End-to-end NGS pipelines: WGS/WES variant calling, RNA-seq DE, scRNA-seq, and metagenomics
- Statistical analysis suite: distribution fitting, GLMs, Bayesian estimation, survival analysis, PCA, and clustering
- Protein thermostability: Tm prediction and mutation design with bioactivity-preserving constraints
- Public archive access: SRA, GEO, ENA, GSA, NCBI Datasets, and GDC
- NCBI E-utilities and RPS-BLAST skills with manuscript-style documentation
54 skills
2 agents
3 manuscripts
Research Infrastructure
Research Knowledge Substrate
An agent-first local research graph system for ingesting papers, extracting structured research objects, querying evidence, and serving a queryable local workspace. Packaged for PyPI distribution.
- Ingest PDFs, DOI, arXiv, PMID, and canonical paper URLs with automatic PDF acquisition
- Extract claims, methods, datasets, and concepts; build a typed graph of evidence relations
- Hybrid lexical/semantic search with local embeddings
- Agent-facing CLI and HTTP service with bundled skill export for Codex and Claude Code
SQLite graph store
CLI + HTTP API
PyPI package
AI Reading Interface
Agent PDF Workbench
An agent-aware PDF workspace that exposes paper reading sessions, structured user interaction events, and durable annotations to AI agents via MCP. Bridges the gap between passive PDF reading and agent-grounded literature review.
- Local PDF.js viewer with event capture: page changes, highlights, copies, and comments
- Durable SQLite event store and annotation/note CRUD — replayable across conversations
- MCP server entrypoint for Codex and Claude Code tool integration
- React + TypeScript frontend; Python backend; full E2E test suite
MCP server
local viewer
event store
Publication Layer
Structured Intelligence Manuscript Series
Publication-style web pages that explain why the workflows matter, where operational barriers exist, and how an independent AI interface layer can improve reuse without hiding the original method.
- Web editions for NCBI E-utilities, RPS-BLAST, and NGS Analysis Expert
- Availability, limitations, and positioning written for scientists
- Markdown sources preserved for citation and revision
web manuscripts
markdown source
availability notes
Publishing Model
The platform is meant to host reusable scientific practice, not just describe it.
What gets published
Skills, scripts, references, manuscripts, validation rules, and reproducible output expectations. The goal is to make scientific operations public enough to inspect and teach.
What remains explicit
This is an independent, unofficial AI interaction layer around upstream tools and services. It does not replace original software, databases, or expert judgment, and it does not imply endorsement.
Organization
Scientific Tooling is now the public-facing layer for the work.
The domain and organization name support a broader initiative now spanning three active repositories: Structured Intelligence for skill-based scientific workflows, Research Knowledge Substrate for local paper graph management, and Agent PDF Workbench for agent-aware literature reading.
The broader initiative adds an independent AI interface layer around existing scientific tools and services. It does not claim ownership of upstream software, databases, or documentation, and it does not imply endorsement by the original authors or institutions.