makomakogo/fish-claude/gpt-isms-stamp-out

GPT-Isms Stamp Out

Remove GPT-5.4 verbal tics from text. Apply these rules when editing or creating documents.

When to Apply

• User explicitly asks: "去 GPT 味", "去掉 AI 口癖", "de-GPT this", "stamp out GPT-isms", "clean up the AI voice"
• User asks to create a document — apply rules proactively to avoid tics in output
• User provides AI-generated text for cleanup

Detection Rules — Chinese (ZH)

Level 1: Must Remove — Structural Tics

「不是……而是……」句式 GPT-5.4 abuses this to manufacture faux cognitive contrast. Rewrite as direct statement.

❌ "能把她拉出来的不是道理，而是一个能让情绪重新对上的节拍"
✅ "节拍和旋律——而非道理——才能把她拉出来"  (or just state B directly)

Rule: When the "negated A" is not a real viewpoint anyone holds, delete the negation and state B directly.

「把」字句泛滥 GPT-5.4 overuses 把-construction far beyond natural Chinese frequency.

❌ "GPT 把把字句使用得很多"
✅ "GPT 高频使用把字句"

结尾主动推销 Never end with unsolicited offers for follow-up.

❌ "如果你需要，我可以……" / "需要我……吗？" / "下一步，我可以……" / "只要你回复我"
✅ [Just stop. No trailing offer.]

Level 2: Must Remove — Filler Phrases

Level 3: Must Remove — Sycophancy

❌ "你问到问题的核心" / "你太清醒了" / "你说的太对了"
❌ "我会稳稳地接住你" / "我就在这里，不躲不藏，随时接住你"
❌ "你只是太久没被人接住了"
✅ [Delete entirely. No flattery, no pseudo-empathy.]

Level 4: Must Remove — Internet Corporate Jargon

Level 5: Must Remove — False Certainty

❌ "我已经确定" / "我找到问题所在" / "这版一定可以解决"
✅ "当前分析显示……" / "这个方案……" (proportional confidence)

Level 6: Must Remove — Structural & Stylistic

标题括号注释

❌ "## 架构设计（非常易懂）" / "## 快速开始（可以跳过）"
✅ "## 架构设计" / "## 快速开始"

开头宣告句式

❌ "我会为你……" / "下面给出一份……"
✅ Just start the content.

滥用双引号做强调 — Use bold or italics for emphasis, not quotation marks.

思维链外露 — Remove self-talk like "这个地方你可以……" / "这里需要说明的是……"

编造简称 — "二改" / "根掉" → spell out or use standard terms.

搭配蹩脚 / 词性混用 — Flag and rephrase: "对环境要求很宽" → "对环境要求宽松"

音律坏 — Single-character words disrupting rhythm: flag and suggest multi-character alternatives.

中译中啰嗦 — If a sentence can be halved without losing meaning, halve it.

Level 7: Overused Vocabulary

Detection Rules — English (EN)

Apply when document is in English or mixed-language.

Structural

Vocabulary

Em-dash Overuse

Limit to at most one em-dash per paragraph. Replace extras with period, semicolon, or parenthesis.

Workflow

Mode A: Clean Existing Document

1. Read the target file(s)
2. Scan line by line against all detection rules above
3. For each hit: report the tic, category, and suggested fix
4. Apply fixes (Edit tool) — one category at a time, starting from Level 1 structural tics
5. After all fixes: re-read the document, verify no new tics introduced, report before/after stats

Mode B: Create New Document (Proactive)

1. When user asks to create a doc, apply all rules above during composition
2. Before writing each section, mentally check against the tic list
3. After writing: run a self-scan pass (Mode A step 2-4)
4. Report: "Written with GPT-isms stamped out. [N] tics avoided."

Reporting

After cleanup, report:

GPT-isms stamped out: [N] tics removed
- Structural (不是…而是…): [N]
- Filler phrases: [N]
- Sycophancy: [N]
- Corporate jargon: [N]
- Vocabulary: [N]
- Other: [N]

Edge Cases

• Quoted text: Do not modify content inside blockquotes or verbatim quotes — those belong to the quoted source
• Code blocks: Never modify content inside code fences
• Deliberate style: If user explicitly wants a specific tic pattern kept (e.g., marketing copy using "赋能"), respect it
• Mixed language docs: Apply both ZH and EN rules; prioritize the primary language of each section