$ show K-Dense-AI/scientific-agent-skills
This repository is a curated collection of 177+ domain-specific scientific 'skills' for AI agents, following the Agent Skills standard. Each skill is a self-contained module providing a CLI tool, documentation, and tests for a specific scientific or technical task—ranging from single-cell RNA-seq analysis (scanpy) and molecular docking (diffdock) to clinical report validation and ISO standards readiness. The architecture is a monorepo of independent skills, each with its own scripts/, references/, assets/, and tests/. Skills are designed to be safe, bounded, and reproducible: they validate inputs before execution, enforce resource limits, avoid network access unless explicitly required, and produce structured JSON reports. The repository also includes a shared test infrastructure (skill_contract) and a meta-skill (autoskill) that can automatically propose new skills from screen activity. While primarily an application library for AI agents, it has strong data/research characteristics due to its heavy use of scientific data formats and validation of research artifacts.
The LLM identified 5 files worth highlighting for reuse:
Converts JATS XML (the standard format for scientific articles) into sectioned plain text. Handles edge cases like restricted articles (where NCBI returns HTTP 200 with an error body), metadata-only documents, and rate limiting. Uses a block/skip tag strategy to preserve reading order and parses XML comments because ElementTree drops them. The script is self-contained, has no internal imports beyond standard library and lxml, and could be extracted into any project needing JATS-to-text conversion.
Generates a standalone SHAP report for a fixed Random Forest model on the breast cancer dataset. Demonstrates non-obvious skill by validating SHAP additivity (using np.testing.assert_allclose), rejecting non-finite values, and producing structured JSON/PNG outputs. The script is fully self-contained (imports only shap, numpy, matplotlib, and standard library) and could be extracted as a template for any tabular model explanation pipeline.
A shared test infrastructure module that provides reusable test factories for validating CLI help output and schematic generation. The cli.help_test_case function automatically verifies that every skill's CLI produces consistent --help output without performing any work. The schematic.review_parsing_test_case validates that review-parsing contracts are correctly implemented. This module is loaded by file location to avoid namespace conflicts and could be extracted into any project needing standardized CLI contract testing.
Implements a meta-skill pipeline that automatically proposes new skills from screen activity. It captures events via screenpipe, clusters them, matches against existing skills using embeddings, and drafts a new skill proposal using an LLM. The script demonstrates a non-obvious pattern of using AI to extend the skill library itself. It is self-contained with its own backend abstraction and configuration, and could be extracted as a template for any system that needs to generate reusable tool definitions from user behavior.
A general-purpose exploratory data analysis tool that handles multiple data formats (CSV, JSON, HDF5, images, sequences) and produces a structured JSON report with distribution summaries, missingness analysis, and data quality warnings. It enforces bounded input sizes, rejects symlinks, and never makes network calls. The script is self-contained (imports only numpy, pandas, and standard library) and could be extracted into any project needing a safe, reproducible EDA step.