Skill UI

This skill facilitates the extraction of population-scale (cohort-level) clinical text from Electronic Health Records (EHR) or FHIR data warehouses using the FHIR Bulk Data Access ($export) operation. It manages the asynchronous job kickoff, polling, and downloading of NDJSON files. The resulting data stream is designed to be fed in batches into OpenMed for mandatory de-identification and Named Entity Recognition (NER) analysis.

This skill loads and parses a structured handoff document, enabling the user to resume work from a previous session. It systematically extracts the project's scope, files involved, completed tasks, key discoveries, and next steps. Crucially, it summarizes its understanding and mandates user confirmation before taking any action, ensuring accurate context and continuity.

This skill manages the loading and efficient caching of OpenMed clinical/biomedical Named Entity Recognition (NER) models. Models are downloaded once from the Hugging Face Hub, enabling full offline operation afterward. It supports model reuse across multiple calls, configuring cache directories, specifying devices (CPU/GPU), and handling various model identifiers (registry keys, full HF IDs, or local paths).

This skill generates browser-playable video assets for web pages by acting as an adapter for the OpenRouter API. It manages the entire asynchronous video generation lifecycle, including submitting the job, polling the status, downloading the final MP4 file, and outputting a structured media manifest for web integration. Use this when a local web video asset is required and configuration details are provided.

This skill is designed to read data records from a specified source file, performing necessary transformations and cleaning to prepare the data for downstream use. It handles the core Extract, Transform, Load (ETL) process, making it ideal for data pipelines, data validation, and preparing raw data into a structured format.

This skill scans a specified local directory, extracts essential metadata from the files, and builds a comprehensive, searchable index. It then securely uploads this generated index to a configured remote search service via HTTP, making local files discoverable across remote platforms.

This command-line tool allows users to upload local files (screenshots, images, PDFs, videos, etc.) directly to GitHub's internal user-attachments endpoint. It outputs a special URL that GitHub automatically renders when pasted into PRs, issues, or comments, facilitating the easy sharing of media and screenshots within development workflows. It also supports downloading existing private attachments.

A utility designed to read the contents of local data files by specifying their file path. This tool is useful for quick data inspection, initial data loading, or preliminary analysis when a user needs to access raw information from various file formats (e.g., TXT, CSV).

This methodology guides users to ask critical, load-bearing questions when faced with ambiguous, vague, or assumption-laden requirements. It helps surface hidden requirements and undefined terms, ensuring that the development process is based on explicit, agreed-upon facts rather than guesswork. Use it to define the true problem and scope before committing to a solution.

This methodology allows engineers and analysts to conduct controlled counterfactual simulations when real-world testing (such as catastrophic failures, extreme load, or irreversible changes) is too rare, costly, or impossible. By isolating a single variable and tracing the mechanistic chain step-by-step, it helps surface critical system breakpoints, invariants, and necessary controls before deployment.