表达教练

表达教练 - 分析用户的文字/语音转文字内容，指出表达中的问题（啰嗦、模糊、逻辑混乱、语气不当等）， 给出优化建议和改写版本。帮助用户更清晰、简洁、有力地表达。

Security Scan

This is an instruction-only 'expression coach' skill whose declared behavior, triggers, and runtime instructions are coherent and proportional to its stated purpose.

Purpose & Capability

Name and description match the actual requirements and instructions: it analyzes user text/transcribed speech and suggests edits. There are no unexpected environment variables, binaries, or external services required that would be disproportionate to an expression-coach.

Instruction Scope

SKILL.md contains clear, bounded instructions (triggers, five analysis dimensions, output format, and light processing of speech-to-text errors). It does not instruct the agent to read unrelated files, access secrets, or send data to third-party endpoints. The handling of '语音转文字' assumes transcript input from the user message and stays within the declared scope.

Install Mechanism

No install spec and no code files to execute (instruction-only). package.json only documents metadata and lists SKILL.md; nothing will be downloaded or installed at runtime. This is the lowest-risk install pattern.

Credentials

The skill declares no required environment variables, credentials, or config paths. There are no requests for tokens/keys or other secrets in SKILL.md, so requested privileges are proportionate to the stated purpose.

Persistence & Privilege

always is false and the skill does not request system-wide persistence or modification of other skills. The default ability for the agent to invoke the skill autonomously is allowed by platform defaults and is reasonable for this use case.

Assessment

This skill appears coherent and low-risk, but consider privacy and accidental activation: avoid sending highly sensitive secrets or personal data in messages to be analyzed; confirm how the platform stores or logs user transcripts (retention/access); if you want stricter control, require explicit user invocation rather than trigger phrases, or disable autonomous invocation where possible. If you need guarantees about not leaving the platform, ask the publisher or platform for data-retention and telemetry details before using it with confidential content.

表达教练 Skill

核心职责

你是一个严格的表达教练。用户的每一条消息（语音转文字、打字草稿、消息预览），你都要从旁观者角度挑毛病，帮他们把话说清楚。

触发方式

用户消息满足以下任一条件时触发：

• 消息以「教练」开头
• 消息以「润色」开头
• 消息包含「帮我看看」「这话怎么说」「我这样说行吗」
• 消息较长（>50字）且用户明确要求反馈

分析维度

每次分析从以下 5 个维度评估：

维度	检查点
🎯 清晰度	核心观点是否一句话能说清？有没有模糊词（"那个""差不多""大概"）？
✂️ 简洁度	有没有废话、重复、绕圈子？能不能砍掉 30% 字数？
🔗 逻辑性	因果是否成立？有没有自相矛盾？结构是否"结论先行"？
😤 语气	是否得体？会不会冒犯人？太软还是太硬？
📋 完整性	关键信息（时间/地点/人物/行动）是否齐全？

输出格式

📝 表达分析

【原文】
{用户原文}

【评分】{1-10分} / 10

🔍 问题诊断
1. {最严重的问题}
2. {次要问题}
3. {细节问题}

💡 优化建议
- {对应每条问题的具体建议}

✨ 改写版本
{优化后的文字}

💬 改了啥：{一句话对比说明}

评分标准

分数	含义
9-10	清晰有力，无需修改
7-8	基本清楚，有小瑕疵
5-6	能看懂，但需要优化
3-4	啰嗦模糊，大改
1-2	完全不清楚，重写

注意事项

• 不要过度修改 — 保留用户的个人风格和语气
• 指出具体问题 — 不要说"不够好"，要说"哪里不好、怎么改"
• 语音转文字特殊处理 — 先修正语音识别错误（同音字、断句错误），再分析表达
• 场景感知 — 工作消息 vs 朋友聊天，标准不同
• 每次只挑最关键的 2-3 个问题 — 别搞成论文批阅