"Router for local-first PII and PHI extraction, de-identification,
Scanned 9/8/2026
Install to Claude Code
npx -y skills add VectorSpaceLab/AREX-Skill --skill deidentification-privacy --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Deidentification Privacy?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/vectorspacelab-deidentification-privacy)More formats (shields.io, HTML) on the badges page.
---
name: deidentification-privacy
description: "Router for local-first PII and PHI extraction, de-identification,
reversible re-identification, date shifting, surrogate generation,
streaming/batch redaction, and audit-safe outputs."
disable-model-invocation: true
metadata:
disco-role: operating
license: Apache 2.0
---
# De-identification and privacy
Use this sub-skill for local-first PII/PHI workflows that need:
- `extract_pii` spans, labels, offsets, and confidences;
- `deidentify` text redaction with masking, removal, replacement, hashing, date shifting, or format-preserving output;
- `reidentify` when an authorized mapping is available;
- policy profiles, surrogate generation, patient-keyed date shifts, audit-safe outputs, multilingual or code-mixed identifiers, and no-PHI logging;
- document, streaming, and batch redaction over synthetic or authorized text.
## Go here first
- Need entity interpretation, model selection, or clinical grounding? Use sibling `clinical-extraction-grounding`.
- Need table/row release risk or structured de-identification? Use sibling `structured-risk-evaluation`.
- Need OCR or document intake before redaction? Use sibling `multimodal-document-intake`.
- Need service, adapter, or FHIR/OMOP handoff? Use sibling `interoperability-serving`.
## Core surfaces
- Top-level APIs: `extract_pii`, `deidentify`, `reidentify`
- Privacy helpers: `openmed.core.anonymizer`, `openmed.core.date_shift`, `openmed.core.surrogate_vault`
- Long-text helpers: `openmed.core.document_stream`, `openmed.core.streaming`
- Batch helpers: `openmed.processing.BatchProcessor`
## Default operating rules
- Keep inputs synthetic or already authorized.
- Prefer local loaders, cached checkpoints, and explicit policy names.
- Never log raw source text, mappings, prompts, or PHI-bearing outputs.
- If the task needs re-identification, treat the mapping as sensitive as the source.
- For long documents, stream or batch rather than loading everything into memory.
## Bundled references
- `references/privacy-workflows.md`
- `references/api-reference.md`
- `references/troubleshooting.md`
- `scripts/deidentify_synthetic_note.py`
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!