Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add Haochenhust/ch-skills --skill meta-promptgit clone --depth 1 https://github.com/Haochenhust/ch-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/haochenhust/ch-skills/meta-prompt)<a href="https://agentmods.dev/skills/haochenhust/ch-skills/meta-prompt"><img src="https://agentmods.dev/badge/skills/haochenhust/ch-skills/meta-prompt/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/haochenhust/ch-skills/meta-prompt"><img src="https://agentmods.dev/badge/skills/haochenhust/ch-skills/meta-prompt.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00144 | $0.03283 |
| Opus 5 | $0.00072 | $0.01641 |
| Sonnet 5 | $0.00029 | $0.00657 |
| Haiku 4.5 | $0.00014 | $0.00328 |
Grade A, and why
meta-prompt scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 254 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Meta-Prompt Skill
定位与边界
做什么：接收用户一段原始任务描述（可以是一句话到一大段话），通过最少的澄清问题 + 按需叠加的 prompt engineering 技巧，输出一段完整、结构化、可直接复制到其他 AI 会话使用的高质量 prompt。
不做什么（严格）：
- 不执行产出的 prompt。用户要的是 prompt 文本本身，不是任务结果。哪怕产出的 prompt 看起来很简单、哪怕你以为顺手做掉更贴心——都不要做。这是这个 skill 最容易踩的坑。
- 不做面向产品的 prompt 工程（安全护栏、A/B 评估、版本管理、一致性测试都不在范围内）。
- 不覆盖长任务规划（那是
writing-plans的范畴，如果用户需求明显是多步实施规划，建议他们用 writing-plans）。
为什么：用户常常在对话里"不得要领"——AI 理解不到重点、格式不对、视角不对。根因 90% 是上下文缺失和约束模糊，不是模型不够强。这个 skill 把"帮用户把任务表达清楚"这件事做好，剩下的交给目标 AI。
全程中文
与用户的所有交互（澄清问题的措辞、技巧说明、交付说明）都用中文。产出的 prompt 本身按目标 AI 使用场景的自然语言——用户任务描述是中文就用中文写 prompt，是英文就用英文。
工作流（严格按顺序）
Step 1 — 意图解析
读入用户原始输入，按以下 7 维度扫描，心里给每一维标记 ✓ 已知 / ✗ 缺失 / ⚠️ 歧义：
| 维度 | 判断要点 |
|---|---|
| 目标 | 用户最终想拿到什么可交付物？ |
| 受众/使用场景 | 产物给谁看、用在什么场景？ |
| 输入数据或上下文 | 有没有素材要喂给目标 AI？（文本、数据、代码…） |
| 约束条件 | 字数、风格、语言、禁用项、必须包含项 |
| 输出格式 | 文本/JSON/列表/表格/代码/markdown 分节… |
| 成功标准 | 什么样算好？用户隐含的质量门槛 |
| 目标 AI | Claude（默认）/ GPT / Gemini / 其他 |
为什么这 7 维：这是勾勒一个任务的最小充分集。少一维 prompt 就会漏信息，多一维就是过度工程。
Step 2 — 动态澄清（用 AskUserQuestion）
只问 ✗ 缺失 / ⚠️ 歧义的维度，一次性批量用 AskUserQuestion 工具提出：
- 问题数量：1–4 个。超过 4 说明你还没把原始输入榨干，先回 Step 1 再读一遍。
- 每题给 3–4 个具体、可选中的选项 + "其他（请说明）"。不要给"详细"/"简洁"这种形容词级选项，要给"<200 字简报"/"300-500 字"/"1 页以上详细"这种有量化锚点的选项。
- 不要问已经明确的维度。用户输入里"写个 Python 函数"已经暗示代码输出格式了，就别再问。
- 原始输入已经非常清晰时（7 维里至少 5 维 ✓），直接跳过 Step 2。
为什么澄清而不改写：用户的日常痛点是"AI 没理解我"，90% 是上下文缺失，10% 是表达问题。澄清直接补上下文，改写只在语言层面化妆。
Step 3 — 技巧选择
根据任务画像从下列菜单挑选技巧。不是越多越好——每个技巧都要能说出"为什么选它"。索引如下，详细模板和触发条件见 references/techniques.md（需要具体模板片段时再读）：
| 技巧 | 一句话触发条件 |
|---|---|
| Role assignment | 任务从特定领域视角做会明显更好（技术/法律/编辑/教师等） |
| Chain-of-Thought | 涉及多步推理、计算、决策、因果分析 |
| Few-shot examples | 格式/风格很特殊，光描述说不清，样例能一秒说清 |
| Self-consistency | 答案方差大的开放题，让模型生成 N 个候选再挑 |
| Step-back prompting | 需要先抽象原则再套具体案例，避免模型一头扎进细节 |
| Chain-of-Verification | 事实性输出，容易幻觉的话题（历史、数据、引用） |
| Decomposition | 任务天然有多个子可交付物，拆开回答质量更高 |
| XML 结构化标签 | Claude 默认开启，多字段输入、长上下文时尤其关键 |
| Markdown 分节 | GPT 默认用，GPT 对 markdown 标题更敏感 |
| Output schema | 产出要被程序解析（JSON、CSV、固定表格结构） |
| Negative constraints | 有明确的失败模式要避免（"不要 xxx"） |
| Delimiter 隔离 | 用户数据里可能混入貌似指令的内容 |
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 254 lines · 144 tokens per session scan A 126069371eda
meta-prompt is a skill published in the GitHub repository Haochenhust/ch-skills (9 stars, last pushed 2mo ago), licensed MIT. It adds 144 tokens to every session and 3,283 once invoked, about $0.0007 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
prompt-master
Generates optimized prompts for AI tools. Activates only when the user explicitly asks to write, fix, improve, or adapt a prompt for a specific AI tool (LLM, Cursor, Midjourney, image AI, video AI, coding agents, etc.). Does not activate for general conversation, coding tasks, document writing, or other…
Prompt Refiner
Improves AI prompts to be clearer, more specific, and produce more consistent outputs.
thanos
Turn any raw prompt into an optimized prompt with the right agent loop, skills, and model. Advisory only — never executes the task itself. TRIGGER when: user invokes /thanos followed by a raw prompt, or says "optimize this prompt", "what loop does this need", "improve my prompt", "how should I prompt for", "rewrite…
encargo
Use when a raw prompt needs rewriting or sharpening, when the user says "rewrite this prompt", "make this prompt better", or asks for a prompt to hand to another agent. Also use before launching any autonomous, long-running, parallel or expensive run (/loop, subagent fan-out, worktrees, overnight work), and whenever a…
prompt-coach
A prompt-writing guide that turns a vague request into a clear goal, its fixed limits, and the main obstacle. It then produces a structured prompt that an AI can act on.
seedance-cinematic-film
Write photorealistic live-action cinematic Seedance 2.0 prompts for Higgsfield, built on five grounding pillars that stop AI drift and floaty motion. Use for "cinematic film prompt", "shot like a movie", realistic body movement, grounded motion, restrained emotional close-ups, driving scenes, fight choreography…