humanize-it

Humanize-It: Iterative de-AI-flavor rewriting

Rewrite the specified document to remove the AI flavor. Based on the text type and content, automatically select the most suitable one from the three sub-skills to begin rewriting. If the rewriting effect is unsatisfactory, switch to another skill and continue iterating until the effect meets the standard or the upper limit of 42 iterations is reached.

Sub-Skill Capability Matrix

Skill	Applicable Scenarios	Core Capabilities	Rewriting Style
humanizer-zh	General text, articles, blogs, copywriting	Identify 24+ AI writing patterns, inject personality and soul, vary rhythm	Natural, warm, with opinions
humanize-chinese	General + academic + long text	20+ rule detection + statistical features + LR fusion scoring, CLI toolchain	Multiple styles available (Zhihu/Xiaohongshu/academic/literary, etc.)
technical-writing	Technical documents, architecture descriptions, design drafts, review documents	Remove technical jargon, evidence first, eliminate host-like tone	Plain, rigorous, arguable

Workflow

Step 1: Read and analyze the document

Read the document content specified by the user
Determine the document type:
- Technical documents(technical proposals, architecture descriptions, design documents, review drafts) → prioritize using technical-writing
- Academic papers(papers, research articles) → prioritize using humanize-chinese(academic mode)
- General text(blogs, copywriting, articles) → prioritize using humanizer-zh
- Long text(≥1500 characters) → prioritize using humanize-chinese(long text mode)

Step 2: Rewriting strategy selection

Select the first rewriting skill based on the document type:

文档类型判断：
├── 技术文档 → technical-writing → humanizer-zh → humanize-chinese
├── 学术论文 → humanize-chinese(academic) → humanizer-zh → technical-writing
├── 通用文本 → humanizer-zh → humanize-chinese → technical-writing
└── 长文本(≥1500字) → humanize-chinese(longform) → humanizer-zh → technical-writing

Step 3: Iterative rewriting loop

iteration = 0
MAX_ITERATIONS = 42

while iteration < MAX_ITERATIONS:
    iteration += 1

    # 使用当前 skill 进行改写
    result = humanize(current_text, current_skill)

    # 评估改写效果
    score = evaluate(result)

    if score >= PASS_THRESHOLD:
        # 改写效果达标，输出最终结果
        output(result)
        break

    # 效果不达标，切换到下一个 skill
    current_skill = next_skill(skill_order)
    current_text = result  # 在上次改写基础上继续优化

    if all_skills_exhausted():
        # 所有 skill 都用过一轮，从头开始新一轮组合
        current_skill = first_skill(skill_order)

Step 4: Quality assessment

After each rewriting, evaluate the effect according to the following dimensions (on a 100-point scale):

Dimension	Weight	Evaluation Criteria
AI trace removal	30%	Whether three-part structures, clichés, and mechanical connectives are eliminated
Naturalness	25%	Whether it reads like it was written by a human, and whether the rhythm is natural
Information completeness	20%	Whether the core information is retained, without losing key content
Style consistency	15%	Whether the tone is consistent throughout and fits the document type
Readability	10%	Whether the sentences are smooth and the logic is clear

Scoring criteria:

≥ 80 points: pass, output the result
60-79 points: acceptable, iterate one more round to see if it can be improved
< 60 points: does not meet the standard, must continue rewriting

Step 5: Output the result

Output the final rewriting result, along with:

The complete rewritten text (written to the original file or a specified output file)
Rewrite summary (which skills were used, how many iterations, final score)
List of main changes

Specific execution instructions

When invoking humanizer-zh

Use the Skill tool to invoke humanizer-zh, pass in the document content, and have it detect and rewrite according to 24 AI writing patterns. Focus on:

Remove filler phrases
Break formulaic structures
Vary rhythm
Trust the reader
Remove punchy one-liners
Inject personality and soul

After rewriting, conduct a preliminary evaluation according to its 50-point quality scoring system (≥ 40 points is considered passing).

When invoking humanize-chinese

Use the Skill tool to invoke humanize-chinese, and select the corresponding mode according to the document type:

General text: use the detect + rewrite process
Academic papers: use academic mode
Long text: use --scene novel or --scene auto

If the CLI tool is available, prioritize using the CLI for quantitative detection and rewriting; otherwise follow its LLM usage guide to execute manually.

After rewriting, evaluate according to its 0-100 scoring system (≤ 35 points is the LOW range, considered passing).

When invoking technical-writing

Use the Skill tool to invoke technical-writing, mainly used for rewriting technical documents. Focus on:

Remove words from the buzzword blacklist
Eliminate host-like and evaluative tone
Rewrite sentences according to "condition → object → judgment"
Evidence first, reduce unsupported strong judgments
Remove transitional sentences

Iteration strategy details

When to switch Skills

The same skill has been used for 3 consecutive rewrites with no score improvement → switch to the next skill
Score decreases → roll back to the previous version and switch skills
All skills have been used for one round but the standard has not been met → start again from the first skill, but use the previous round's rewrite result as input
Reach 42 iterations → output the current best result

Combination strategy

Different skills have different rewriting focuses, and using them in combination can complement each other:

humanizer-zh → humanize-chinese: first inject personality and soul, then perform quantitative de-AI-ification
humanize-chinese → humanizer-zh: first perform systematic de-AI-ification, then inject warmth
technical-writing → humanizer-zh: first ensure technical rigor, then increase naturalness
humanizer-zh → technical-writing: first remove general AI traces, then adjust the technical tone

User interaction

If the user provides explicit preferences (such as "keep the academic style", "make it more colloquial", "technical documents should not be too casual"), follow the user's preferences when rewriting, and prioritize the corresponding skill.

If the user does not specify the document type, you can ask via AskUserQuestion:

Document type (general / technical / academic)
Desired rewriting style (natural and casual / plain and rigorous / academic standard)
Whether it needs to be written to the original file or output to a new file