Skill UI

A comprehensive guide for auditing research papers submitted to the WSDM conference. It details critical constraints, including the no-rebuttal policy, strict 9-page budgets (including appendices), and stringent hybrid anonymization requirements. Use this guide to preemptively identify and correct submission flaws before the deadline.

Provides a comprehensive, quantitative auditing system for AI agents, skills, and commands. It evaluates these components against industry best practices using a multi-criteria weighted scoring model and assign a production readiness grade (A-F). Ideal for validating complex AI systems before deployment, identifying gaps, and ensuring robust, maintainable code.

This skill performs an adversarial second pass to scan already de-identified clinical text for residual identifiers (PHI/PII). It combines deterministic regex checks (SSN, phone, dates) with a model-based PII detection rerun. It scores severity and implements a hard block rule: any detected leak prevents data release, ensuring the highest standard of data privacy and compliance.

Generates a signed, reproducible, and non-PHI audit trail for de-identification processes. This skill provides compliance evidence, including offsets, content hashes, and model provenance, enabling users to verify that sensitive identifiers were correctly redacted for HIPAA/GDPR audits without exposing any plaintext PHI. It is crucial for defensible data handling.

Designed for regulated environments (GxP/GCP), this skill generates and verifies audit trails compliant with 21 CFR Part 11 standards. It ensures every data processing step leaves an attributable, time-stamped, and tamper-evident record, providing electronic signatures and robust data integrity checks (ALCOA+). It is essential for clinical trials and regulated data governance.

This skill provides a comprehensive checklist for auditing de-identified Protected Health Information (PHI) against the 18 categories required by HIPAA Safe Harbor (45 CFR 164.514(b)(2)). It maps detected PII entities to these categories, identifies gaps (e.g., age > 89, rare geography), and guides the assessment of residual re-identification risk. This is essential for proving the completeness and compliance of de-identification efforts.

This skill audits OpenMed NER or de-identification models to detect performance disparities across sensitive demographic subgroups (e.g., sex, age, race). It calculates metrics like per-group recall and leakage disparity, helping users identify instances of 'under-protection' before deployment, ensuring equitable AI governance and compliance.

This tool benchmarks the recall of PII detection models, crucial for ensuring privacy compliance before deployment. It uses synthetic fixtures and enforces a recall floor by measuring label-aware exact-span and grapheme recall against predefined gold standards. It is essential for auditing models and preventing critical privacy leaks.

This guide provides a comprehensive methodology for the reverse engineering of browser extensions (Chrome/Firefox). It covers analyzing the manifest, tracing background workers, content scripts, and recovering sensitive logic like credentials or network traffic. Essential for security research, malware investigation, and supply chain security auditing.

This skill acts as a comprehensive map for developing clinical and biomedical NLP pipelines using OpenMed. OpenMed is a local-first Python library that performs named-entity recognition (NER), protected health information (PHI) de-identification, and FHIR export, all running fully on-device without cloud calls. Use this guide when starting a new clinical text project, needing to select the appropriate capability (e.g., zero-shot NER, auditing, or FHIR conversion), or requiring an entry point for the core OpenMed features.

A comprehensive white-box security auditing workflow designed for proprietary source code. It integrates static analysis security testing (SAST) tools like Semgrep and CodeQL to detect common vulnerabilities (e.g., injection, IDOR, crypto misuse), followed by mandatory manual threat modeling and verification steps. Applicable for PR reviews and pre-deployment security checks.

A specialized skill for analyzing and reproducing complex, timing-dependent bugs in concurrent systems. Use this to investigate race windows, ordering bugs, lock gaps, stale reads, or state inconsistencies where the final outcome depends critically on request timing or asynchronous execution order. Essential for deep security auditing and debugging of distributed transactions.