| name | moai-domain-humanize |
| description | AI text humanization and 윤문 (post-editing) specialist that detects and removes AI tells while preserving meaning, facts, and figures. Covers Korean, English, Japanese, and Chinese with a shared severity model (S1/S2/S3), quality grades (A/B/C/D), and 30%/50% over-editing guardrails. Use to make AI-generated text read as human-authored without changing what it says (de-ai, naturalness pass).
|
| user-invocable | false |
| version | 3.1.0 |
moai-domain-humanize
Post-editing specialist that removes "AI tells" from generated text and rewrites it to read as human-authored, while preserving meaning. This is the editing counterpart to text generation: it does not write new content, it refines how existing content is said. Covers Korean, English, Japanese, and Chinese.
Quick Reference
Operating Principles (4)
- Meaning preservation is the top rule. Facts, numbers, statistics, named entities, quotations, citations, and the author's stance/certainty stay intact. Any meaning drift forces a rollback.
- Evidence-based edits only. Every change must trace to a detected tell on a specific span. Stylistic "improvements" unconnected to a catalogued tell are themselves an over-editing signal and are forbidden.
- Genre and register preservation. Humanize within the source register — academic stays academic, casual stays casual. Never push formal text into slang or vice versa.
- Over-editing prevention. Flag at >30% change (WARN), halt at >50% change (forced stop / human review). Above 50% you are regenerating, not humanizing.
Mode Selection
- Fast mode (default, up to ~5,000 chars): a single pass — detect, rewrite, self-verify against the meaning-preservation checklist.
- Strict mode (long or high-stakes text, or when requested): separate stages — detect → surgical rewrite → content-fidelity audit (facts/figures/stance unchanged) → naturalness review. Re-run a second pass when the result lands at Grade C.
Output Contract
Return two things:
- The humanized text.
- A short change report: categories hit (with counts), the final quality grade (A/B/C/D), and the percent changed (character-level edit distance ÷ source length). When a guardrail fires, state it explicitly (WARN at >30%, HALT at >50%).
Common Severity Model (shared by all 4 languages)
Each tell carries one severity tier. Detectors gate by occurrence count and overlap, because a single tell rarely proves AI authorship — confidence comes from clustering.
| Tier | Name | Rule |
|---|
| S1 | Decisive | A single occurrence strongly confirms AI authorship → remove on first occurrence. |
| S2 | Strong | Acceptable at 1–2 instances → remove at 3 or more. |
| S3 | Weak | Problematic only when overlapping other tells → downgrade-only contributor. |
Common Quality Grades (shared by all 4 languages)
Graded after the rewrite, on the residual S1/S2 counts plus improvement % (= proportion of detected tells removed without introducing new ones).
| Grade | Criteria | Action |
|---|
| A | 0 residual S1, ≤2 residual S2, ≥70% improvement | Pass — reads as human-authored |
| B | 0 residual S1, ≤4 residual S2, ≥50% improvement | Pass — minor polish remains |
| C | 1–2 residual S1, OR <50% improvement, OR over-edit WARN (>30%) | Trigger a second pass |
| D | ≥3 residual S1, OR over-edit HALT (>50%), OR meaning drift detected | Request human review; do not auto-ship |
Hard rule: any residual S1 caps the grade at C; any meaning-distortion flag forces D. S3 tells affect the grade only when ≥3 of them overlap and reinforce an S1/S2 finding.
Over-Editing Guardrails (shared)
Change rate = character-level edit distance ÷ source length; target band ~5–30%.
- >30% changed → WARN. Surface a caution and cap at Grade C until each edit is justified by a detected tell. Note: padding-removal legitimately shrinks text, so a length drop alone is not a violation — flag when meaning-bearing spans are altered.
- >50% changed → HALT. Stop and require human confirmation; revert to the last safe state.
Meaning-Preservation Checklist (shared, all must hold)
- Anchor facts first — fix the claims, numbers, names, dates, and certainty level before editing.
- Edit at sentence/phrase level, not whole-document regeneration.
- Add no new facts — never invent specifics to replace vagueness; simplify instead, or flag for the author.
- Drop no load-bearing facts — removing an inflated wrapper must keep the substantive claim inside.
- Preserve genuine certainty/hedging and technical terminology verbatim.
- Final diff check — compare facts, tone, certainty, and examples against the original; revert any edit that drifts.
Language Routing
Each target language has its own tell catalogue (categories, before/after examples in the target language, per-category severity). Load the module that matches the text being edited:
| Language | Module | Source basis |
|---|
| Korean (한국어) | modules/korean.md | Faithful port of the im-not-ai (Humanize KR) taxonomy (10 categories A–J, 100+ subcategories) |
| English | modules/english.md | Web-researched catalogue (10 categories EN-A … EN-J) |
| Japanese (日本語) | modules/japanese.md | Web-researched catalogue (9 categories JA-01 … JA-09) |
| Chinese (中文) | modules/chinese.md | Web-researched catalogue (11 categories CN-A … CN-K) |
The Korean module is a faithful port of the open-source im-not-ai taxonomy; the English, Japanese, and Chinese modules are independently web-researched catalogues modeled on the same architecture. The common severity model and quality grades above apply uniformly to every module — the modules add only the language-specific tell categories, severities, and example rewrites.
For mixed-language text, detect the dominant language and route to its module; apply each module independently to its spans when the text is genuinely multilingual.
Implementation Guide
Workflow (per text)
- Identify language and mode. Pick the module by dominant language; pick Fast vs Strict by length / stakes.
- Anchor facts. Record the numbers, names, dates, quotations, and stance that must not change (meaning-preservation checklist item 1).
- Detect tells. Scan against the module's catalogue. Record each hit with its category ID, span, and severity. Count occurrences (S2/S3 gate on repetition).
- Rewrite surgically. Edit only flagged spans. Replace each tell with a natural rendering in the same register. Do not touch unflagged text.
- Measure change rate. Compute character-level edit distance ÷ source length. Apply guardrails (WARN >30%, HALT >50%).
- Self-verify (Fast) or audit + review (Strict). Re-run the meaning-preservation checklist. In Strict mode, run the content-fidelity audit and naturalness review as separate stages.
- Grade. Count residual S1/S2 and improvement %; assign A/B/C/D. Second pass on C; human review on D.
- Emit the humanized text + change report.
Detection note (shared across languages)
Automated AI-text detectors are unreliable across these four languages (notably weak on CJK polite registers, where they false-positive on correct formal writing). This skill is a pattern-based editing tool, not a detection oracle: rely on the catalogued tell categories and the clustering-based severity gates, not on a detector's verdict.
Common pitfalls
- Re-injecting AI-ness. Rewriting AI text with a fresh full regeneration tends to add new tells. Favor surgical edits to flagged spans over wholesale rewriting.
- Fabricating specifics. When a module calls for concrete detail to replace vague filler and no real specifics exist, simplify or flag for the author — never invent.
- Style drift. "Cleaning up" beyond the flagged tells violates principle 2 and inflates the change rate toward the HALT threshold.
- Mixing registers mid-document. Keep one consistent register (and, for Japanese/Chinese, one consistent politeness/sentence-ending style) across the whole output.
Works Well With
sync-auditor: independent skeptical review. Use it to score whether the humanized output preserved meaning against the original and met the target grade.
Korean patterns adapted from the im-not-ai (Humanize KR) open-source skill — see NOTICE.md.
Version: 1.0.0
