skills.rest
maziyarpanahi
maziyarpanahiCOMMUNITY
Explore all skills by this author →
Website𝕏@maziyarpanahiGitHub

auditing-deid-leakage

Detect residual identifiers in de-identified clinical text and block release.

5.0k|615|Updated Oct 4, 2025
Search Tags:#de-identification#release gating#leakage detection#clinical text#phi detection#privacy auditing
One-click install
npx skills add https://github.com/maziyarpanahi/openmed --skill auditing-deid-leakage
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: auditing-deid-leakage
Source: https://github.com/maziyarpanahi/openmed/tree/main/skills/auditing-deid-leakage
Command: npx skills add https://github.com/maziyarpanahi/openmed --skill auditing-deid-leakage

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill verifies that clinical text believed to be de-identified contains no residual identifiers before export, sharing, training, or publication. It prevents a single missed SSN, email, phone number, account number, date, or other structured identifier from undermining the entire release.

Core Features & Use Cases

  • Adversarial leakage detection: Scans de-identified text for SSNs, emails, phone numbers, dates, MRNs, account identifiers, and payment-card patterns.
  • Validation and risk scoring: Uses regular expressions, Luhn checksum validation, entropy heuristics, and a second-pass OpenMed PII detector to identify potential leaks and assign severity.
  • Release gating: Produces a privacy-preserving report containing labels, offsets, counts, and severities without echoing leaked plaintext, and blocks release when high- or critical-severity identifiers are found.
  • Use Case: Before publishing a clinical dataset, run this audit on the de-identified output, review any findings, re-redact the source, and repeat the scan until the release gate passes.

Quick Start

Use the auditing-deid-leakage skill to scan the attached de-identified clinical text for residual identifiers and return a leakage report containing only finding labels, severities, offsets, and release status.

Frequently Asked Questions about auditing-deid-leakage

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I detect residual PHI in de-identified clinical text before sharing?▼

Detect residual PHI in de-identified clinical text by applying deterministic format detectors, checksum validation, entropy heuristics, and a second-pass OpenMed PII extraction to block release when high-severity identifiers are found.

What is release gating for clinical datasets and how does it work?▼

Release gating for clinical datasets is a privacy verification process that blocks publication when high-severity residual identifiers are found, producing reports with labels, offsets, counts, and severities while excluding leaked plaintext.

How do I automate privacy auditing for continuous integration checks?▼

Automate privacy auditing for continuous integration checks by running de-identification leakage detection on clinical text outputs to catch privacy regressions, assigning severity scores to identifiers, and blocking deployment when critical leaks are found.

Does this approach validate Luhn checksums and entropy heuristics for PHI detection?▼

This PHI detection approach validates Luhn checksums and entropy heuristics by applying regular expressions and a second-pass OpenMed PII detector to identify structured identifiers like payment cards and assign severity risk scores.

What is the best way to verify de-identified clinical text for publication review?▼

The best way to verify de-identified clinical text for publication review is to run an adversarial leakage audit that scans for structured identifiers, assigns severity scores, and blocks release until all high-severity findings are re-redacted.

Related Skills

Explore More Legal & Compliance Skills→
maziyarpanahimaziyarpanahi
5.0k

auditing-deidentification-runs

Generate signed, PHI-free audit reports for OpenMed de-identification runs.

Community
Advanced
maziyarpanahimaziyarpanahi
5.0k

reviewing-reidentification-risk

Assesses re-identification risk in de-identified datasets using k-anonymity and l-diversity analyses.

Community
Advanced
maziyarpanahimaziyarpanahi
5.0k

choosing-openmed-models

Select OpenMed clinical or biomedical models via structured ModelQuery filters.

Community
Intermediate
AlphaBitCoreAlphaBitCore
21

oss-secret-scan

Scan repositories for leaked secrets, PII, and infrastructure data.

Official
Advanced