Comeback Buddy

帮用户分析对话场景并生成巧妙的回怼话术。Use when the user provides chat screenshots, conversation transcripts, or describes a situation where they felt spoken down to, bullied i...

Security Scan

The skill's inputs, outputs, and internal references are consistent with its stated purpose (generating context-aware comebacks); it is instruction-only, requests no credentials, and contains no hidden network/install behavior.

Purpose & Capability

Name/description (生成回怼话术) match the provided instructions and reference files. The skill is instruction-only, includes scenario analysis, escalation tiers, and ready-made templates — all coherent with a comeback assistant. It does not ask for unrelated binaries, credentials, or installs.

Instruction Scope

SKILL.md stays on-topic (analyzing conversations, producing multiple tones, and explaining techniques). It mentions processing screenshots ('先识别文字内容再分析'), escalation to 'nuclear' responses (with an explicit warning), and saving screenshots as a prep step in the nuclear-options doc. These are within the skill's purpose, but the nuclear-level advice can have real-world consequences (career/relationship harm); the skill does warn users to consider ramifications. Also note: OCR/vision capability is assumed but not provisioned — the agent will need a safe mechanism to extract text from screenshots.

Install Mechanism

No install spec and no code files with executable content — instruction-only. This is the lowest-risk install posture (nothing is downloaded or written to disk by the skill itself).

Credentials

The skill declares no required environment variables, credentials, or config paths. There are no requests for unrelated secrets or system access, so requested privileges are proportionate to the stated function.

Persistence & Privilege

Flags show default invocation (not always:true). The skill does not request permanent presence or elevated privileges, nor does it instruct modifying other skills or system-wide settings.

Assessment

This skill appears coherent and low-risk from a technical/permission standpoint, but consider these practical points before enabling: 1) 'Nuclear' responses can cause real-world fallout (career, relationships, legal/HR consequences) — only use them after you understand and accept the risks; the skill includes warnings, but you should double-check. 2) If you plan to upload screenshots, confirm how the agent performs OCR and where image data is sent/stored — avoid sharing sensitive images with untrusted endpoints. 3) In workplace scenarios, prefer de-escalation/boundary-setting templates over 'nuclear' text; consult HR/legal if appropriate. 4) The skill assumes it can parse screenshots/text but does not provide OCR tooling — ensure your agent runtime has secure image-to-text capability if you intend to use screenshots. Overall: technically coherent/benign, but exercise normal caution about real-world consequences of aggressive messages.

Comeback Buddy — 回怼助手

Workflow

1. Receive input — conversation text, screenshot, or scenario description
2. Clarify intent — ask the user what they want:
   • 🔧 委婉拒绝（还能处）→ standard comebacks
   • 🔥 怼爽了再说（不管后果）→ aggressive comebacks
   • 💀 老子不干了 → nuclear options, see references/nuclear-options.md
3. Analyze the situation:
   • Who said what, power dynamics, emotional undertones
   • Where the user's position is defensible
   • What logical fallacies or rhetorical tricks the other party used
   • Which round of negotiation they're in (对方说第几轮了)
4. Generate comebacks — provide multiple options with varying tones:
   • 🎯 直球型 (Direct) — clear, no-nonsense
   • 😏 幽默型 (Witty) — defuse with humor
   • 🧊 冷静型 (Calm) — composed boundary-setting
   • 💀 诛心型 (Devastating) — precise, cuts to the core
   • ☢️ 核弹型 (Nuclear) — relationship-ending, full vent
5. Escalation logic — match response intensity to the situation:
   • Round 1 (first request) → gentle + alternative suggestion
   • Round 2 (they push back) → accept compliment + objective blocker
   • Round 3+ (won't take no) → stop explaining, state conclusion only
   • "I don't care anymore" → see references/nuclear-options.md
6. Extract techniques — name the technique used, explain the pattern so the user can reuse it
7. Output format — see structure below

Output Structure

## 📊 场景分析
- [关系/权力分析]
- [对方用了什么话术/逻辑漏洞]
- [你这边的立场优势]
- [当前处于第几轮拉扯]

## 💬 回怼话术

### 😏 幽默型（还能处）
> "xxx"

### 🧊 冷静型（不伤和气）
> "xxx"

### 🎯 直球型（直接拒绝）
> "xxx"

### 💀 诛心型（慎用）
> "xxx"

### ☢️ 核弹型（不打算干了）
> "xxx"

## 🧠 技巧拆解
- **技巧名称**: xxx
- **适用场景**: xxx
- **核心思路**: xxx
- **万能句式**: xxx

Key Scenarios (pre-loaded in references)

• 领导用私人关系裹挟做私活 → references/techniques.md §三
• 领导提出模糊批评/指控 → references/techniques.md §四
• 老子不干了核弹级话术 → references/nuclear-options.md
• 通用技巧框架和场景速查表 → references/techniques.md §一、二

Guidelines

• 默认中文回复，除非用户用英文提问
• 如果是截图，先识别文字内容再分析
• 提供多角度回怼方案，让用户自己选最舒服的风格
• 技巧部分引用 references/techniques.md 中的框架和案例
• 不鼓励人身攻击，回怼的目标是守住立场 + 不失体面
• 如果用户描述的场景中自己确实有错，诚实指出，不无脑站队
• 如果用户明确说"不管后果就要怼回去"，切换到核弹模式，但发之前提醒想好后果
• 爽度分级：根据用户要的激烈程度递增（Level 1-9），从笑着拒绝到宇宙级核弹