ai-check

Name: ai-check
Author: harshaneel

45installs31forks

How to install ai-check?

harshaneel/humanize/ai-check

$npx -y skills add harshaneel/humanize --skill ai-check

Installs into the current project.

Summary

Use when someone asks "does this sound AI?", "check if this is AI-written", "what gives this away as AI", "run ai-check on this", or "score this text". Also use when reviewing a draft for AI tells before publishing, or when a piece of text reads as suspiciously polished, generic,

Files · 1

View on GitHub

SKILL.md

1	# AI-Check Skill
2
3	Forensic analysis of text for AI-generation signals. Grounded in the published detection
4	literature (Wu et al. 2025, Mitchell et al. 2023, Kujur 2025, AAAI 2025 shared task).
5
6	The output is a structured report, not a vague judgment. Every fired signal cites evidence.
7
8	---
9
10	## The nine signal categories
11
12	Score each category 0–3:
13	- 0 = No signal detected (human-consistent)
14	- 1 = Weak signal (possible AI, could be human)
15	- 2 = Moderate signal (likely AI pattern)
16	- 3 = Strong signal (near-certain AI pattern)
17
18	Severity-to-score mapping (use for every category):
19
20	\| Evidence in category \| Score \|
21	\|---\|---\|
22	\| No flagged instances \| 0 \|
23	\| One weak instance, or vague unease without a specific quote \| 1 \|
24	\| One moderate instance, or two or more weak instances \| 2 \|
25	\| One strong instance, or two or more moderate instances, or four or more weak instances \| 3 \|
26
27	Double-counting policy: a single phrase can fire at most two distinct signals when the phrase is genuinely diagnostic for both. Example: "it is important to note that" is both Signal A (banned vocabulary) and Signal C (institutional hedge). Log it under both, but the same phrase cannot count as two separate weak instances inside the same category.
28
29	Total score cap: 9 categories × 3 = 27 maximum.
30
31	### Signal A: Perplexity (word predictability)
32
33	Look for vocabulary that is maximally safe and expected — words that are technically correct
34	but never the most precise or interesting choice a knowledgeable human would make.
35
36	Flags:
37	- Generic verbs where domain-specific ones belong ("address" instead of "untangle", "implement" instead of "wire up")
38	- Adjectives that describe without adding information ("significant improvements", "notable progress", "key challenges")
39	- Hedged assertions that swap specificity for safety ("can often lead to", "may result in", "tends to")
40	- Any of the canonical AI vocabulary list:
41	delve, leverage (verb), utilize, robust, comprehensive, streamline, foster, facilitate,
42	pivotal, nuanced, notable, notably, enduring, garner, it is worth noting, it is important to note,
43	multifaceted, in the realm of, the landscape of, a myriad of, a plethora of
44
45	Cite the exact word or phrase that fired.
46
47	### Signal B: Burstiness deficit (sentence uniformity)
48
49	Measure the variation in sentence length across the text.
50
51	Flags:
52	- Three or more consecutive sentences within 5 words of the same length
53	- No sentence shorter than 8 words in any 150-word block
54	- Metronomic rhythm — reading the passage aloud produces a steady pulse rather than natural variation
55	- No fragments used for emphasis
56
57	Report: list the sentence lengths in sequence (e.g. "14, 16, 13, 15, 17 — five consecutive sentences within 4 words of each other").
58
59	### Signal C: Hedge density
60
61	Count the softening and epistemic hedge words.
62
63	Flags:
64	- "often", "generally", "typically", "in many cases", "it can be argued" appearing where direct assertion is warranted
65	- "it is important to note that", "it is worth mentioning", "one might consider"
66	- Diplomatic framing of obvious tradeoffs: "while X has benefits, it also presents challenges"
67	- Uncertainty expressed as institutional hedging rather than personal ("results may vary") vs human ("I'm not sure this holds when...")
68
69	Report: quote each hedge and note whether it was warranted by genuine uncertainty or reflexive softening.
70
71	### Signal D: Structural tells
72
73	Look for document architecture patterns AI imposes regardless of content.
74
75	Flags:
76	- Bullet list where prose would serve better
77	- Topic sentence + evidence + restatement of topic sentence (humans skip the restatement)
78	- "In conclusion / To summarize / In summary" openers on closing paragraphs
79	- "In this [post/article/section] I will..." openers
80	- Numbered steps for content that isn't genuinely sequential
81	- Three-part structure imposed on every paragraph (intro, body, conclusion at micro-scale)
82	- Tricolon parallel structure: three examples or beats with identical grammatical shape
83	e.g. "You X. Y. Does Z? You X. Y. Does Z? You X. Y. Does Z?" — perfectly symmetrical
84	triplets in prose are AI-constructed. Real writers use two examples or vary the shape.
85	Severity: strong.
86	- Perfect paragraph-per-idea arc: every paragraph does exactly one narrative job and
87	advances the arc cleanly (setup → tension → lesson → evidence → reflection). Real personal
88	writing has a paragraph that meanders, does two jobs, or doesn't fully resolve. A piece
89	where every paragraph lands cleanly is architecturally perfect in a way human writing
90	isn't. Severity: moderate in isolation, strong combined with othe

People who liked this also viewed

6 picks

Type

ai-check

45installs31forks

View on GitHub

How to install ai-check?

harshaneel/humanize/ai-check

$npx -y skills add harshaneel/humanize --skill ai-check

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# AI-Check Skill
2
3	Forensic analysis of text for AI-generation signals. Grounded in the published detection
4	literature (Wu et al. 2025, Mitchell et al. 2023, Kujur 2025, AAAI 2025 shared task).
5
6	The output is a structured report, not a vague judgment. Every fired signal cites evidence.
7
8	---
9
10	## The nine signal categories
11
12	Score each category 0–3:
13	- 0 = No signal detected (human-consistent)
14	- 1 = Weak signal (possible AI, could be human)
15	- 2 = Moderate signal (likely AI pattern)
16	- 3 = Strong signal (near-certain AI pattern)
17
18	Severity-to-score mapping (use for every category):
19
20	\| Evidence in category \| Score \|
21	\|---\|---\|
22	\| No flagged instances \| 0 \|
23	\| One weak instance, or vague unease without a specific quote \| 1 \|
24	\| One moderate instance, or two or more weak instances \| 2 \|
25	\| One strong instance, or two or more moderate instances, or four or more weak instances \| 3 \|
26
27	Double-counting policy: a single phrase can fire at most two distinct signals when the phrase is genuinely diagnostic for both. Example: "it is important to note that" is both Signal A (banned vocabulary) and Signal C (institutional hedge). Log it under both, but the same phrase cannot count as two separate weak instances inside the same category.
28
29	Total score cap: 9 categories × 3 = 27 maximum.
30
31	### Signal A: Perplexity (word predictability)
32
33	Look for vocabulary that is maximally safe and expected — words that are technically correct
34	but never the most precise or interesting choice a knowledgeable human would make.
35
36	Flags:
37	- Generic verbs where domain-specific ones belong ("address" instead of "untangle", "implement" instead of "wire up")
38	- Adjectives that describe without adding information ("significant improvements", "notable progress", "key challenges")
39	- Hedged assertions that swap specificity for safety ("can often lead to", "may result in", "tends to")
40	- Any of the canonical AI vocabulary list:
41	delve, leverage (verb), utilize, robust, comprehensive, streamline, foster, facilitate,
42	pivotal, nuanced, notable, notably, enduring, garner, it is worth noting, it is important to note,
43	multifaceted, in the realm of, the landscape of, a myriad of, a plethora of
44
45	Cite the exact word or phrase that fired.
46
47	### Signal B: Burstiness deficit (sentence uniformity)
48
49	Measure the variation in sentence length across the text.
50
51	Flags:
52	- Three or more consecutive sentences within 5 words of the same length
53	- No sentence shorter than 8 words in any 150-word block
54	- Metronomic rhythm — reading the passage aloud produces a steady pulse rather than natural variation
55	- No fragments used for emphasis
56
57	Report: list the sentence lengths in sequence (e.g. "14, 16, 13, 15, 17 — five consecutive sentences within 4 words of each other").
58
59	### Signal C: Hedge density
60
61	Count the softening and epistemic hedge words.
62
63	Flags:
64	- "often", "generally", "typically", "in many cases", "it can be argued" appearing where direct assertion is warranted
65	- "it is important to note that", "it is worth mentioning", "one might consider"
66	- Diplomatic framing of obvious tradeoffs: "while X has benefits, it also presents challenges"
67	- Uncertainty expressed as institutional hedging rather than personal ("results may vary") vs human ("I'm not sure this holds when...")
68
69	Report: quote each hedge and note whether it was warranted by genuine uncertainty or reflexive softening.
70
71	### Signal D: Structural tells
72
73	Look for document architecture patterns AI imposes regardless of content.
74
75	Flags:
76	- Bullet list where prose would serve better
77	- Topic sentence + evidence + restatement of topic sentence (humans skip the restatement)
78	- "In conclusion / To summarize / In summary" openers on closing paragraphs
79	- "In this [post/article/section] I will..." openers
80	- Numbered steps for content that isn't genuinely sequential
81	- Three-part structure imposed on every paragraph (intro, body, conclusion at micro-scale)
82	- Tricolon parallel structure: three examples or beats with identical grammatical shape
83	e.g. "You X. Y. Does Z? You X. Y. Does Z? You X. Y. Does Z?" — perfectly symmetrical
84	triplets in prose are AI-constructed. Real writers use two examples or vary the shape.
85	Severity: strong.
86	- Perfect paragraph-per-idea arc: every paragraph does exactly one narrative job and
87	advances the arc cleanly (setup → tension → lesson → evidence → reflection). Real personal
88	writing has a paragraph that meanders, does two jobs, or doesn't fully resolve. A piece
89	where every paragraph lands cleanly is architecturally perfect in a way human writing
90	isn't. Severity: moderate in isolation, strong combined with othe

People who liked this also viewed

6 picks

Type