SciTeX Skills (scitex-skills)
Aggregated AI-agent skill pages for the SciTeX ecosystem — 230+ markdown trigger pages, hourly auto-refreshed from PyPI wheels.
Docs ·
scitex.ai ·
pip install scitex-skills
Part of the SciTeX ecosystem -- structured skill pages that help AI agents (Claude Code, MCP clients) discover and use SciTeX packages. Each skill is a focused Markdown document covering one capability -- API signatures, CLI commands, MCP tools, and usage patterns -- so agents can look up exactly what they need without reading source code.
Problem and Solution
| # | Problem | Solution |
|---|---|---|
| 1 | AI agents must read source code to learn each package's API -- costly tokens, stale inference, no trigger phrases | pip install scitex-skills -- 130+ curated skill pages with frontmatter triggers; Claude Code auto-invokes the right package without reading src/ |
| 2 | Per-package _skills/ directories are scattered across 33 repos -- discovery, search, and browse are fragmented |
One aggregated tree at skills/<pkg>/ -- one pip install, one mkdocs site, one symlink target for ~/.claude/skills/ |
| 3 | Per-release skill drift -- documentation in published wheels goes stale between releases | Hourly GitHub Actions refresh from PyPI -- aggregator is never more than ~60 min behind the latest released wheels |
Features
- 130+ skill pages across 15+ packages -- from statistics to manuscript compilation
- Auto-updated hourly from PyPI via GitHub Actions -- always ≤60 min behind released versions
- Four-interface coverage -- every skill documents Python API, CLI, MCP tools, and skill discovery
- Structured for AI agents -- concise, machine-readable Markdown with frontmatter metadata
- Browsable documentation -- hosted on Read the Docs
Skill Packages
233 skill pages across 32 packages (regenerated hourly from PyPI wheels)
| Package | Skills | Description |
|---|---|---|
| crossref-local | 16 | Local CrossRef database with 167M+ works and full-text search |
| figrecipe | 11 | Reproducible matplotlib figures with mm-precision layouts, declarative specs, YA |
| general | 16 | SciTeX ecosystem general standards — branding, package architecture, four interf |
| scitex-agent-container | 6 | Declarative YAML-based AI agent lifecycle management with tmux/screen, SSH remot |
| scitex-app | 12 | App developer SDK — scaffold, validate, dev-install, standalone shell, file oper |
| scitex-audio | 6 | Text-to-speech with multiple backends (ElevenLabs, LuxTTS, gTTS, pyttsx3), smart |
| scitex-audit | 4 | Unified repo security scanner for scientific Python projects — one call orchestr |
| scitex-browser | 7 | Playwright wrappers for scientific web scraping + AI-agent browsing — adds debug |
| scitex-clew | 4 | Hash-based reproducibility verification for scientific pipelines |
| scitex-compat | 2 | Backward-compatibility shims for deprecated SciTeX APIs — holds legacy aliases s |
| scitex-container | 5 | Unified container management for Apptainer/Singularity and Docker — build SIFs |
| scitex-core | 4 | Foundation layer for the SciTeX ecosystem — imported by every other SciTeX packa |
| scitex-dataset | 4 | Unified dataset-discovery API across 7 scientific repositories — OpenNeuro + DAN |
| scitex-db | 6 | Relational-DB wrapper for scientific Python — SQLite3 and PostgreSQL classes |
| scitex-dev | 12 | Shared developer utilities for the SciTeX ecosystem — ecosystem management, vers |
| scitex-dict | 2 | Dictionary utilities for scientific Python — attribute-access dicts, conflict-aw |
| scitex-etc | 2 | Miscellaneous SciTeX utilities — terminal single-keypress helpers for interactiv |
| scitex-gists | 2 | SigmaPlot v12 macro snippets as printable Python functions — for researchers who |
| scitex-io | 8 | Universal file I/O supporting 30+ formats (CSV, NumPy, pickle, YAML, JSON, HDF5 |
| scitex-linter | 4 | AST-based linter for reproducible-research Python — 47 built-in rules across 7 c |
| scitex-logging | 2 | Enhanced Python logging + warnings + exceptions for SciTeX — stdlib-compatible w |
| scitex-orochi | 44 | Agent Communication Hub — real-time WebSocket messaging between AI agents across |
| scitex-parallel | 1 | Minimal thread-pool parallel execution for scientific Python — a single one-shot |
| scitex-path | 2 | Project-aware path utilities for scientific Python — finding files/dirs/git root |
| scitex-repro | 2 | Reproducibility helpers for scientific Python experiments — unified seeded rando |
| scitex-scholar | 11 | Scientific-paper search, metadata enrichment, PDF download, and BibTeX library m |
| scitex-stats | 6 | Publication-ready statistical testing — 23 tests, effect sizes, power analysis |
| scitex-str | 2 | Text-processing utilities for scientific Python — colored/boxed console prints |
| scitex-tunnel | 5 | Persistent SSH reverse tunnel for NAT traversal - auto-reconnecting tunnels for |
| scitex-types | 2 | Type aliases and runtime type guards for scientific Python — unions across NumPy |
| scitex-ui | 6 | Vanilla TS workspace shell framework + React app components for SciTeX apps |
| scitex-writer | 17 | LaTeX manuscript compilation with bibliography, figures, tables, claims tracking |
Installation
pip install scitex-skills
1 Interfaces
Skills (Markdown for AI agents)
Skills are bundled as package data. AI agents and tools can query them programmatically:
from scitex_skills import get_skill, list_skills
# List all available skills
skills = list_skills()
# Get a specific skill page
content = get_skill("scitex-io/save-and-load")
Browse online at [scitex-skills.readthedocs.io](https://scitex-skills.readthedocs.io/)
or as files under [`skills/`](skills/).
How It Works
An hourly GitHub Actions workflow keeps skills in sync with released packages:
- Installs
scitex-devfrom PyPI - Calls
export_skills(source="pypi")to download all SciTeX wheels - Extracts
_skills/Markdown files from each wheel - Commits any changes to
skills/
Manual Update
pip install scitex-dev
python -c "from pathlib import Path; from scitex_dev.skills import export_skills; export_skills(source='pypi', dest=Path('skills'))"
Browse Skills
- Online: scitex-skills.readthedocs.io
- Local: See the
skills/directory
To build the docs locally:
pip install mkdocs
mkdocs serve
Contributing
Skill pages are auto-generated from each SciTeX package. To contribute:
- Add or edit
_skills/*.mdfiles in the source package (e.g.,scitex-io,figrecipe) - Release a new version of that package to PyPI
- The hourly workflow will pick up the changes automatically (within 60 min)
For ecosystem-wide standards, edit files under skills/scitex/general/.
License
AGPL-3.0 -- see LICENSE for details.
Part of SciTeX
scitex-skills is part of SciTeX. Install via
the umbrella with pip install scitex[skills] to use as
scitex.skills (Python) or scitex skills ... (CLI).
Four Freedoms for Research
- The freedom to run your research anywhere — your machine, your terms.
- The freedom to study how every step works — from raw data to final manuscript.
- The freedom to redistribute your workflows, not just your papers.
- The freedom to modify any module and share improvements with the community.
AGPL-3.0 — because we believe research infrastructure deserves the same freedoms as the software it runs on.