humanizer

Name: humanizer
Author: Borda

23stars3forks

How to install humanizer?

borda/ai-rig/humanizer

$npx -y skills add Borda/AI-Rig --skill humanizer

Installs into the current project.

Summary

Strip AI-writing tells from prose destined for humans — docs, PR/commit bodies, reports, release notes, blog posts, Slack/email drafts. Removes LLM-vocabulary clichés (delve, boasts, testament, underscore, robust, tapestry...), banned constructions (not just X but Y, rule-of-thre

Files · 1

View on GitHub

SKILL.md

1	<objective>
2
3	Detect and remove statistical AI-writing fingerprints from human-facing prose before it ships. Grounded in Wikipedia's crowd-sourced AI-detection corpus (`Wikipedia:Signs of AI writing`) — a maintained list of vocabulary, syntax, and formatting patterns that over-represent in LLM output vs human baseline. Apply as a final pass, not a rewrite-from-scratch: preserve meaning, facts, and structure; only excise the tells.
4
5	</objective>
6
7	<inputs>
8
9	- text or file path to humanize: optional. Inline text, or a file path (Markdown/plain text) to edit in place.
10	- check `<file>`: read-only mode — report findings without editing.
11	- No argument: humanize the draft already composed earlier in this turn (self-review pass) — only reachable when the model chooses to invoke this skill mid-task; there is no platform hook that guarantees a pre-send interception, so treat this path as best-effort, not a hard gate.
12
13	</inputs>
14
15	<workflow>
16
17	Task hygiene: call `TaskList` first; triage orphaned tasks. Task tracking: skip for single-pass humanize calls under 3 steps; use for multi-file batch runs.
18
19	## 1. Load the target text
20
21	- Inline text → work on it directly, no file I/O.
22	- File path → `Read` the file.
23	- No argument → treat the draft already composed earlier in this turn as the target.
24
25	## 2. Scan against the checklist
26
27	Walk the text once per category below; flag every hit before editing anything (report-first, matches `check` mode output).
28
29	Vocabulary — cut or replace with plain equivalent:
30
31	\| Banned \| Plain replacement \|
32	\| --- \| --- \|
33	\| delve, boasts, testament, underscore(s), showcase, tapestry, intricate/intricacies, meticulous, robust, vibrant, pivotal, crucial, garner, foster(ing), align with, landscape, interplay, enduring, enhance \| say the specific thing instead — drop the word, don't swap in another vague one \|
34	\| "stands as", "serves as", "marks a", "represents" (as copula dodge) \| "is" / "was" \|
35	\| "Additionally,", "Moreover,", "It is important to note that" \| delete, or state the fact directly \|
36
37	Syntax — flag and restructure:
38
39	- Negative parallelism: "not just X, but Y" / "not X, but Y" / "not only X but also Y" / "X rather than Y" used as a crutch
40	- Rule-of-three triads used for false comprehensiveness ("fast, reliable, and scalable")
41	- "-ing" superficial-analysis tails: "highlighting...", "underscoring...", "contributing to..." tacked onto a claim with no source
42	- Vague attribution / weasel words: "industry reports", "observers", "experts argue", "some critics" with no named source
43	- Formulaic "Despite its [positives], X faces challenges..." conclusion pattern
44
45	Formatting — flag and fix:
46
47	- Title Case In Headings → sentence case
48	- Mechanical bolding of every instance of a repeated term
49	- Markdown overuse — bold/bullets/headers where a plain sentence reads fine; the single most common tell in PR bodies and reports
50	- Bare-bullet inline-header lists (`• Header: text`) where prose or a real table reads better
51	- Em dash overuse — chain of `—` clauses instead of periods/commas
52	- Curly ("smart") quotes/apostrophes mixed inconsistently with straight ones
53	- `---`/`***` thematic breaks before headings (Markdown artifact bleeding into prose)
54
55	## 3. Apply fixes
56
57	- `check` mode: stop here — report findings (category, location, quote, suggested fix), do not edit.
58	- Edit mode: apply the minimal edit per flagged instance using `Edit`. Preserve every fact, number, and citation — only the phrasing/formatting changes. Re-read the result once to confirm no fact was dropped in the rewrite.
59
60	## 4. Report
61
62	One line per category with hit count and net edits made (e.g. "vocabulary: 4 removed, syntax: 2 restructured, formatting: 1 fixed"). Zero hits → say so plainly, do not pad the report.
63
64	</workflow>

People who liked this also viewed

6 picks

Type

humanizer

23stars3forks

View on GitHub

How to install humanizer?

borda/ai-rig/humanizer

$npx -y skills add Borda/AI-Rig --skill humanizer

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	<objective>
2
3	Detect and remove statistical AI-writing fingerprints from human-facing prose before it ships. Grounded in Wikipedia's crowd-sourced AI-detection corpus (`Wikipedia:Signs of AI writing`) — a maintained list of vocabulary, syntax, and formatting patterns that over-represent in LLM output vs human baseline. Apply as a final pass, not a rewrite-from-scratch: preserve meaning, facts, and structure; only excise the tells.
4
5	</objective>
6
7	<inputs>
8
9	- text or file path to humanize: optional. Inline text, or a file path (Markdown/plain text) to edit in place.
10	- check `<file>`: read-only mode — report findings without editing.
11	- No argument: humanize the draft already composed earlier in this turn (self-review pass) — only reachable when the model chooses to invoke this skill mid-task; there is no platform hook that guarantees a pre-send interception, so treat this path as best-effort, not a hard gate.
12
13	</inputs>
14
15	<workflow>
16
17	Task hygiene: call `TaskList` first; triage orphaned tasks. Task tracking: skip for single-pass humanize calls under 3 steps; use for multi-file batch runs.
18
19	## 1. Load the target text
20
21	- Inline text → work on it directly, no file I/O.
22	- File path → `Read` the file.
23	- No argument → treat the draft already composed earlier in this turn as the target.
24
25	## 2. Scan against the checklist
26
27	Walk the text once per category below; flag every hit before editing anything (report-first, matches `check` mode output).
28
29	Vocabulary — cut or replace with plain equivalent:
30
31	\| Banned \| Plain replacement \|
32	\| --- \| --- \|
33	\| delve, boasts, testament, underscore(s), showcase, tapestry, intricate/intricacies, meticulous, robust, vibrant, pivotal, crucial, garner, foster(ing), align with, landscape, interplay, enduring, enhance \| say the specific thing instead — drop the word, don't swap in another vague one \|
34	\| "stands as", "serves as", "marks a", "represents" (as copula dodge) \| "is" / "was" \|
35	\| "Additionally,", "Moreover,", "It is important to note that" \| delete, or state the fact directly \|
36
37	Syntax — flag and restructure:
38
39	- Negative parallelism: "not just X, but Y" / "not X, but Y" / "not only X but also Y" / "X rather than Y" used as a crutch
40	- Rule-of-three triads used for false comprehensiveness ("fast, reliable, and scalable")
41	- "-ing" superficial-analysis tails: "highlighting...", "underscoring...", "contributing to..." tacked onto a claim with no source
42	- Vague attribution / weasel words: "industry reports", "observers", "experts argue", "some critics" with no named source
43	- Formulaic "Despite its [positives], X faces challenges..." conclusion pattern
44
45	Formatting — flag and fix:
46
47	- Title Case In Headings → sentence case
48	- Mechanical bolding of every instance of a repeated term
49	- Markdown overuse — bold/bullets/headers where a plain sentence reads fine; the single most common tell in PR bodies and reports
50	- Bare-bullet inline-header lists (`• Header: text`) where prose or a real table reads better
51	- Em dash overuse — chain of `—` clauses instead of periods/commas
52	- Curly ("smart") quotes/apostrophes mixed inconsistently with straight ones
53	- `---`/`***` thematic breaks before headings (Markdown artifact bleeding into prose)
54
55	## 3. Apply fixes
56
57	- `check` mode: stop here — report findings (category, location, quote, suggested fix), do not edit.
58	- Edit mode: apply the minimal edit per flagged instance using `Edit`. Preserve every fact, number, and citation — only the phrasing/formatting changes. Re-read the result once to confirm no fact was dropped in the rewrite.
59
60	## 4. Report
61
62	One line per category with hit count and net edits made (e.g. "vocabulary: 4 removed, syntax: 2 restructured, formatting: 1 fixed"). Zero hits → say so plainly, do not pad the report.
63
64	</workflow>

People who liked this also viewed

6 picks

Type