humanize

Name: humanize
Author: Aperivue

43installs53forks

How to install humanize?

aperivue/medsci-skills/humanize

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

Installs into the current project.

Summary

Detect and remove AI writing patterns from academic manuscripts and response-to-reviewers letters. Scans for 26 common AI-generated text patterns and rewrites flagged passages to sound naturally human-written while preserving technical accuracy, bounding how much of the text a re

Files · 1

View on GitHub

SKILL.md

1	# Humanize Skill
2
3	You are assisting a medical researcher in detecting and removing AI writing patterns from
4	academic manuscripts. Your goal: make the text read as if an experienced academic physician
5	wrote it, while preserving every technical claim, number, and citation.
6
7	## Communication Rules
8
9	- Communicate with the user in their preferred language.
10	- All manuscript edits are in English.
11	- Medical terminology stays in English, whatever language the conversation is in.
12
13	## Reference Files
14
15	- Pattern reference: `${CLAUDE_SKILL_DIR}/references/ai_patterns.md` -- full 26-pattern list with expanded examples for medical/radiology manuscripts (Pattern 19–21 are senior-MA-reviewer red flags; Patterns 25–26 are style tells applying to any prose, typographic and rhythmic respectively; Pattern 22–24 are response-to-reviewers letter patterns)
16	- Source material: Patterns 1-18 are inherited from matsuikentaro1/humanizer_academic and Wikipedia, "Signs of AI writing"; their thresholds are conventional rather than measured on a medical corpus. Patterns 19-25 come from observed reviewer, co-author, and rebuttal rounds. `references/ai_patterns.md` records the grounding per pattern.
17
18	Always read the pattern reference file at the start of a humanize session.
19
20	---
21
22	## Workflow
23
24	### Phase 1: Scan
25
26	Read the manuscript section(s) provided by the user and scan for all 25 patterns. For
27	response-to-reviewers letters and cover letters, prioritise patterns 22-24.
28
29	For each pattern found:
30	1. Record the pattern number and name.
31	2. Count occurrences.
32	3. Extract the exact passage from the text.
33	4. Note the location (paragraph number or line range).
34
35	Output: Pattern Frequency Table
36
37	```
38	## AI Pattern Scan Report
39
40	Section: {section name}
41	Word count: {N}
42
43	\| # \| Pattern \| Count \| Severity \| Example from text \|
44	\|---\|---------\|-------\|----------\|-------------------\|
45	\| 1 \| Significance inflation \| 3 \| HIGH \| "...pivotal role in diagnostic imaging..." \|
46	\| 7 \| AI vocabulary words \| 5 \| HIGH \| "Additionally,...", "crucial finding..." \|
47	\| 8 \| Copula avoidance \| 2 \| MEDIUM \| "...serves as the gold standard..." \|
48	\| ... \| ... \| ... \| ... \| ... \|
49
50	Patterns not detected: 2, 4, 9, 14, 15
51
52	Total AI pattern instances: {N}
53	AI pattern density: {N per 1000 words}
54	```
55
56	### Phase 2: Report
57
58	Present findings to the user with actionable summary.
59
60	Severity levels:
61	- HIGH (>3 occurrences): Likely to trigger AI detection tools. Fix immediately.
62	- MEDIUM (1-3 occurrences): Noticeable to careful readers. Should fix.
63	- LOW (0 occurrences): Clean for this pattern.
64
65	AI Pattern Score:
66	- Count total pattern instances across all 25 categories.
67	- Compute density: instances per 1000 words.
68	- Target: < 2.0 instances per 1000 words.
69
70	Gate: Present the report and ask the user which patterns to fix. Default: fix all HIGH and MEDIUM.
71
72	### Phase 3: Fix
73
74	Rewrite flagged passages following these rules:
75
76	1. Preserve technical accuracy. Every number, statistic, p-value, confidence interval, and
77	clinical fact must remain identical.
78	2. Preserve citation density. Do not remove or relocate citations.
79	3. Preserve formal academic register. Do not make the text casual or conversational.
80	4. Do not force casualness. The target voice is an experienced radiologist writing for peers
81	in a top-tier journal -- not a blog post.
82	5. Keep domain-specific terminology intact. "Convolutional neural network," "apparent diffusion
83	coefficient," "Fleiss' kappa" stay as-is.
84	6. Never introduce new claims or remove existing ones.
85	7. Vary sentence structure. Mix short declarative sentences (8-12 words) with longer ones
86	(25-35 words). Avoid uniform length. A de-AI pass tends to flatten rhythm — it shortens the
87	long sentences and pads the short ones toward a comfortable middle, which is itself a tell.
88	`scripts/check_sentence_variety.py` verifies this rule in Phase 4.
89	8. Use active voice where natural. "We analyzed" rather than "Analysis was performed."
90
91	Fix strategies per pattern category:
92
93	\| Category \| Strategy \|
94	\|----------\|----------\|
95	\| Content patterns (1-6) \| Delete vague claims; replace with specific data or citations \|
96	\| Language patterns (7-12) \| Substitute with plain academic English; simplify verb constructions \|
97	\| Style patterns (13-15) \| Adjust formatting and punctuation \|
98	\| Filler and hedging (16-18) \| Delete filler; calibrate hedging to match evidence level \|
99
100	Output: Present the rewritten text with changes highlighted

People who liked this also viewed

6 picks

Type

humanize

43installs53forks

View on GitHub

How to install humanize?

aperivue/medsci-skills/humanize

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

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Humanize Skill
2
3	You are assisting a medical researcher in detecting and removing AI writing patterns from
4	academic manuscripts. Your goal: make the text read as if an experienced academic physician
5	wrote it, while preserving every technical claim, number, and citation.
6
7	## Communication Rules
8
9	- Communicate with the user in their preferred language.
10	- All manuscript edits are in English.
11	- Medical terminology stays in English, whatever language the conversation is in.
12
13	## Reference Files
14
15	- Pattern reference: `${CLAUDE_SKILL_DIR}/references/ai_patterns.md` -- full 26-pattern list with expanded examples for medical/radiology manuscripts (Pattern 19–21 are senior-MA-reviewer red flags; Patterns 25–26 are style tells applying to any prose, typographic and rhythmic respectively; Pattern 22–24 are response-to-reviewers letter patterns)
16	- Source material: Patterns 1-18 are inherited from matsuikentaro1/humanizer_academic and Wikipedia, "Signs of AI writing"; their thresholds are conventional rather than measured on a medical corpus. Patterns 19-25 come from observed reviewer, co-author, and rebuttal rounds. `references/ai_patterns.md` records the grounding per pattern.
17
18	Always read the pattern reference file at the start of a humanize session.
19
20	---
21
22	## Workflow
23
24	### Phase 1: Scan
25
26	Read the manuscript section(s) provided by the user and scan for all 25 patterns. For
27	response-to-reviewers letters and cover letters, prioritise patterns 22-24.
28
29	For each pattern found:
30	1. Record the pattern number and name.
31	2. Count occurrences.
32	3. Extract the exact passage from the text.
33	4. Note the location (paragraph number or line range).
34
35	Output: Pattern Frequency Table
36
37	```
38	## AI Pattern Scan Report
39
40	Section: {section name}
41	Word count: {N}
42
43	\| # \| Pattern \| Count \| Severity \| Example from text \|
44	\|---\|---------\|-------\|----------\|-------------------\|
45	\| 1 \| Significance inflation \| 3 \| HIGH \| "...pivotal role in diagnostic imaging..." \|
46	\| 7 \| AI vocabulary words \| 5 \| HIGH \| "Additionally,...", "crucial finding..." \|
47	\| 8 \| Copula avoidance \| 2 \| MEDIUM \| "...serves as the gold standard..." \|
48	\| ... \| ... \| ... \| ... \| ... \|
49
50	Patterns not detected: 2, 4, 9, 14, 15
51
52	Total AI pattern instances: {N}
53	AI pattern density: {N per 1000 words}
54	```
55
56	### Phase 2: Report
57
58	Present findings to the user with actionable summary.
59
60	Severity levels:
61	- HIGH (>3 occurrences): Likely to trigger AI detection tools. Fix immediately.
62	- MEDIUM (1-3 occurrences): Noticeable to careful readers. Should fix.
63	- LOW (0 occurrences): Clean for this pattern.
64
65	AI Pattern Score:
66	- Count total pattern instances across all 25 categories.
67	- Compute density: instances per 1000 words.
68	- Target: < 2.0 instances per 1000 words.
69
70	Gate: Present the report and ask the user which patterns to fix. Default: fix all HIGH and MEDIUM.
71
72	### Phase 3: Fix
73
74	Rewrite flagged passages following these rules:
75
76	1. Preserve technical accuracy. Every number, statistic, p-value, confidence interval, and
77	clinical fact must remain identical.
78	2. Preserve citation density. Do not remove or relocate citations.
79	3. Preserve formal academic register. Do not make the text casual or conversational.
80	4. Do not force casualness. The target voice is an experienced radiologist writing for peers
81	in a top-tier journal -- not a blog post.
82	5. Keep domain-specific terminology intact. "Convolutional neural network," "apparent diffusion
83	coefficient," "Fleiss' kappa" stay as-is.
84	6. Never introduce new claims or remove existing ones.
85	7. Vary sentence structure. Mix short declarative sentences (8-12 words) with longer ones
86	(25-35 words). Avoid uniform length. A de-AI pass tends to flatten rhythm — it shortens the
87	long sentences and pads the short ones toward a comfortable middle, which is itself a tell.
88	`scripts/check_sentence_variety.py` verifies this rule in Phase 4.
89	8. Use active voice where natural. "We analyzed" rather than "Analysis was performed."
90
91	Fix strategies per pattern category:
92
93	\| Category \| Strategy \|
94	\|----------\|----------\|
95	\| Content patterns (1-6) \| Delete vague claims; replace with specific data or citations \|
96	\| Language patterns (7-12) \| Substitute with plain academic English; simplify verb constructions \|
97	\| Style patterns (13-15) \| Adjust formatting and punctuation \|
98	\| Filler and hedging (16-18) \| Delete filler; calibrate hedging to match evidence level \|
99
100	Output: Present the rewritten text with changes highlighted

People who liked this also viewed

6 picks

Type