deidentify

Name: deidentify
Author: Aperivue

208stars53forks

How to install deidentify?

aperivue/medsci-skills/deidentify

$npx -y skills add Aperivue/medsci-skills --skill deidentify

Installs into the current project.

Summary

De-identify clinical research data before LLM-assisted analysis. Standalone Python CLI detects PHI via regex + heuristics with 10 country locale packs (kr, us, jp, cn, de, uk, fr, ca, au, in). Interactive terminal review. No LLM touches raw data — the script runs locally without

Files · 1

View on GitHub

SKILL.md

1	# De-identification Skill
2
3	You are guiding a medical researcher through data de-identification. The actual
4	de-identification is performed by a standalone Python script that runs WITHOUT
5	any LLM. Your role is to explain, guide, and verify — not to see or process raw
6	PHI data.
7
8	## Critical Safety Rules
9
10	1. NEVER ask the user to paste, show, or upload raw data containing PHI.
11	The script processes data locally. You never need to see patient-level data.
12	2. NEVER read or display the mapping file contents. It contains original PHI values.
13	3. You may read the scan report (column classifications, no raw values), audit log
14	(SHA-256 hashes only), and de-identified output (PHI already removed).
15	4. Always communicate in the user's preferred language about the process, but use
16	English for technical terms (PHI, HIPAA, Safe Harbor, etc.).
17
18	## Reference Files
19
20	- `${CLAUDE_SKILL_DIR}/references/hipaa_18_identifiers.md` — HIPAA Safe Harbor checklist
21	- `${CLAUDE_SKILL_DIR}/references/korean_phi_patterns.md` — Korean-specific regex patterns
22	- `${CLAUDE_SKILL_DIR}/references/date_shift_guide.md` — Date shifting best practices
23
24	Read relevant references before advising the researcher.
25
26	## Prerequisites
27
28	- Python 3.10+
29	- `openpyxl` (for .xlsx files): `pip install openpyxl`
30	- Supported formats: CSV, TSV, Excel (.xlsx)
31
32	## Five-Phase Workflow
33
34	### Phase 1: Assessment
35
36	Ask the researcher:
37	1. What file format is the data? (CSV, Excel, etc.)
38	2. What PHI do you expect in the data? (names, dates, IDs, etc.)
39	3. Does your IRB require specific de-identification documentation?
40	4. Do you need to re-identify later? (affects mapping file choice)
41
42	Based on answers, recommend the appropriate command:
43	- Full pipeline (most common): `python deidentify.py full <file> --locale <code>`
44	- Step-by-step (cautious): `python deidentify.py scan <file> --locale <code>` first
45
46	Available locale codes: `kr` (Korea), `us` (USA), `jp` (Japan), `cn` (China), `de` (Germany),
47	`uk` (United Kingdom), `fr` (France), `ca` (Canada), `au` (Australia), `in` (India).
48	If `--locale` is omitted, the script shows an interactive country selection menu.
49	Users can provide a custom locale file via `--locale-file custom.json`.
50
51	### Phase 2: Script Execution
52
53	Guide the researcher to run the script. The script is located at:
54	```
55	${CLAUDE_SKILL_DIR}/deidentify.py
56	```
57
58	Full pipeline (recommended for most users):
59	```bash
60	python ${CLAUDE_SKILL_DIR}/deidentify.py full data.xlsx \
61	--locale kr \
62	--output-dir ./deidentified/ \
63	--auto-accept-safe
64	```
65
66	Step-by-step (for careful review):
67	```bash
68	# Step 1: Scan
69	python ${CLAUDE_SKILL_DIR}/deidentify.py scan data.xlsx --locale kr --output-dir ./deidentified/
70
71	# Step 2: Review (interactive)
72	python ${CLAUDE_SKILL_DIR}/deidentify.py review ./deidentified/scan_report.json
73
74	# Step 3: Apply
75	python ${CLAUDE_SKILL_DIR}/deidentify.py apply ./deidentified/reviewed_report.json
76	```
77
78	Options:
79	- `--locale CODE`: Country locale for PHI patterns (kr, us, jp, cn, de, uk, fr, ca, au, in)
80	- `--locale-file PATH`: Custom locale JSON file (copy `locales/_template.json` to create one)
81	- `--auto-accept-safe`: Skip confirmation for columns classified as SAFE (faster for large datasets)
82	- `--hash-mapping`: Store SHA-256 hashes instead of original values in mapping file (one-way, more secure)
83	- `--output-dir`: Where to save de-identified file, mapping, and audit log
84	- `-v/--verbose`: Enable debug logging
85
86	### Phase 3: Interactive Review Guidance
87
88	The script's terminal review has three passes:
89
90	1. Pass 1 — Column Classification: Each column is shown as PHI / REVIEW_NEEDED / SAFE.
91	The researcher confirms or overrides each classification.
92	2. Pass 2 — Undecided Items: Columns that weren't resolved in Pass 1 get a second look
93	with more sample values displayed.
94	3. Pass 3 — Final Summary: A table of all planned actions. The researcher can edit
95	individual decisions before confirming.
96
97	Coach the researcher. Deliver these prompts in the researcher's preferred language:
98	- "Columns classified as PHI are anonymized by default. Press 'k' to keep the original value."
99	- "REVIEW_NEEDED are columns the script could not classify. Check the sample values and decide."
100	- "SAFE means no PHI detected. Press 'r' to request re-review if any column looks suspicious."
101
102	### Phase 4: Verify and Document
103
104	After the script completes, help the researcher verify:
105
106	1. Read the audit log (safe — contains only hashes):
107	```bash
108	cat ./deidentified/

People who liked this also viewed

6 picks

Type

deidentify

208stars53forks

View on GitHub

How to install deidentify?

aperivue/medsci-skills/deidentify

$npx -y skills add Aperivue/medsci-skills --skill deidentify

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# De-identification Skill
2
3	You are guiding a medical researcher through data de-identification. The actual
4	de-identification is performed by a standalone Python script that runs WITHOUT
5	any LLM. Your role is to explain, guide, and verify — not to see or process raw
6	PHI data.
7
8	## Critical Safety Rules
9
10	1. NEVER ask the user to paste, show, or upload raw data containing PHI.
11	The script processes data locally. You never need to see patient-level data.
12	2. NEVER read or display the mapping file contents. It contains original PHI values.
13	3. You may read the scan report (column classifications, no raw values), audit log
14	(SHA-256 hashes only), and de-identified output (PHI already removed).
15	4. Always communicate in the user's preferred language about the process, but use
16	English for technical terms (PHI, HIPAA, Safe Harbor, etc.).
17
18	## Reference Files
19
20	- `${CLAUDE_SKILL_DIR}/references/hipaa_18_identifiers.md` — HIPAA Safe Harbor checklist
21	- `${CLAUDE_SKILL_DIR}/references/korean_phi_patterns.md` — Korean-specific regex patterns
22	- `${CLAUDE_SKILL_DIR}/references/date_shift_guide.md` — Date shifting best practices
23
24	Read relevant references before advising the researcher.
25
26	## Prerequisites
27
28	- Python 3.10+
29	- `openpyxl` (for .xlsx files): `pip install openpyxl`
30	- Supported formats: CSV, TSV, Excel (.xlsx)
31
32	## Five-Phase Workflow
33
34	### Phase 1: Assessment
35
36	Ask the researcher:
37	1. What file format is the data? (CSV, Excel, etc.)
38	2. What PHI do you expect in the data? (names, dates, IDs, etc.)
39	3. Does your IRB require specific de-identification documentation?
40	4. Do you need to re-identify later? (affects mapping file choice)
41
42	Based on answers, recommend the appropriate command:
43	- Full pipeline (most common): `python deidentify.py full <file> --locale <code>`
44	- Step-by-step (cautious): `python deidentify.py scan <file> --locale <code>` first
45
46	Available locale codes: `kr` (Korea), `us` (USA), `jp` (Japan), `cn` (China), `de` (Germany),
47	`uk` (United Kingdom), `fr` (France), `ca` (Canada), `au` (Australia), `in` (India).
48	If `--locale` is omitted, the script shows an interactive country selection menu.
49	Users can provide a custom locale file via `--locale-file custom.json`.
50
51	### Phase 2: Script Execution
52
53	Guide the researcher to run the script. The script is located at:
54	```
55	${CLAUDE_SKILL_DIR}/deidentify.py
56	```
57
58	Full pipeline (recommended for most users):
59	```bash
60	python ${CLAUDE_SKILL_DIR}/deidentify.py full data.xlsx \
61	--locale kr \
62	--output-dir ./deidentified/ \
63	--auto-accept-safe
64	```
65
66	Step-by-step (for careful review):
67	```bash
68	# Step 1: Scan
69	python ${CLAUDE_SKILL_DIR}/deidentify.py scan data.xlsx --locale kr --output-dir ./deidentified/
70
71	# Step 2: Review (interactive)
72	python ${CLAUDE_SKILL_DIR}/deidentify.py review ./deidentified/scan_report.json
73
74	# Step 3: Apply
75	python ${CLAUDE_SKILL_DIR}/deidentify.py apply ./deidentified/reviewed_report.json
76	```
77
78	Options:
79	- `--locale CODE`: Country locale for PHI patterns (kr, us, jp, cn, de, uk, fr, ca, au, in)
80	- `--locale-file PATH`: Custom locale JSON file (copy `locales/_template.json` to create one)
81	- `--auto-accept-safe`: Skip confirmation for columns classified as SAFE (faster for large datasets)
82	- `--hash-mapping`: Store SHA-256 hashes instead of original values in mapping file (one-way, more secure)
83	- `--output-dir`: Where to save de-identified file, mapping, and audit log
84	- `-v/--verbose`: Enable debug logging
85
86	### Phase 3: Interactive Review Guidance
87
88	The script's terminal review has three passes:
89
90	1. Pass 1 — Column Classification: Each column is shown as PHI / REVIEW_NEEDED / SAFE.
91	The researcher confirms or overrides each classification.
92	2. Pass 2 — Undecided Items: Columns that weren't resolved in Pass 1 get a second look
93	with more sample values displayed.
94	3. Pass 3 — Final Summary: A table of all planned actions. The researcher can edit
95	individual decisions before confirming.
96
97	Coach the researcher. Deliver these prompts in the researcher's preferred language:
98	- "Columns classified as PHI are anonymized by default. Press 'k' to keep the original value."
99	- "REVIEW_NEEDED are columns the script could not classify. Check the sample values and decide."
100	- "SAFE means no PHI detected. Press 'r' to request re-review if any column looks suspicious."
101
102	### Phase 4: Verify and Document
103
104	After the script completes, help the researcher verify:
105
106	1. Read the audit log (safe — contains only hashes):
107	```bash
108	cat ./deidentified/

People who liked this also viewed

6 picks

Type