gating-deid-leakage

Fail CI when PHI recall drops below approved thresholds or critical identifiers leak.

5.0k|615|Updated Oct 4, 2025
One-click install
npx skills add https://github.com/maziyarpanahi/openmed --skill gating-deid-leakage
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: gating-deid-leakage
Source: https://github.com/maziyarpanahi/openmed/tree/main/skills/gating-deid-leakage
Command: npx skills add https://github.com/maziyarpanahi/openmed --skill gating-deid-leakage

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill prevents weakened clinical de-identification from reaching production by failing CI when PHI recall drops below policy thresholds or critical identifiers leak.

Core Features & Use Cases

  • Recall Regression Gates: Enforce per-label recall floors against a synthetic held-out PHI evaluation set.
  • Zero-Leakage Assertions: Fail closed when any critical identifier, such as an SSN or credit-card number, is detected.
  • CI Integration: Wire pytest or the OpenMed release-gate CLI into GitHub Actions and block merges when evaluation errors, quarantined reports, or failed gates occur.
  • Use Case: Add a required pull-request check that evaluates a candidate de-identification model, compares it with a pinned baseline, and prevents release when leakage or recall regressions are found.

Quick Start

Use the gating-deid-leakage skill to add a required GitHub Actions check that runs the OpenMed de-identification suite and fails on critical leakage or recall below the approved floor.

Frequently Asked Questions about gating-deid-leakage

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

FAQPage Schema
How do I block de-identification regressions in GitHub Actions before merging?

To block de-identification regressions in GitHub Actions, you wire a pytest gate or release-gate CLI into a required pull-request check that fails CI when critical identifiers leak or PHI recall drops below an approved threshold.

What is a recall floor validation for clinical de-identification?

Recall floor validation for clinical de-identification is a release gating mechanism that enforces per-label PHI recall minimums against a held-out synthetic evaluation set, failing closed if the model drops below approved policy thresholds.

How do I set up zero-leakage assertions for PHI in pytest?

Setting up zero-leakage assertions for PHI in pytest involves configuring fail-closed evaluations that immediately fail CI when any critical identifier, such as an SSN or credit-card number, is detected in the de-identified output.

Can I use a pinned baseline model to prevent PHI leakage in CI?

Yes, you can use a pinned baseline model to prevent PHI leakage in CI by comparing candidate de-identification models against the pinned milestone, blocking merges when evaluation errors or recall regressions occur.

Does the OpenMed ReleaseGate integrate with pytest for release gating?

Yes, the OpenMed ReleaseGate integrates with pytest by applying release gating checks that evaluate held-out synthetic PHI, ensuring fail-closed validation and merge-blocking when quarantined reports or failed gates occur.