Product Title Generation

Product Title Generation

Use this skill to turn a product name, feature description, technical capability, platform module, learning service, activity theme, or mixed Chinese-English concept into compact Chinese product titles.

The output is a recommendation plus six alternatives, designed for UI entry names, module names, product cards, page section titles, or brand-like short names.

Ownership

MUST generate short product, feature, module, entry, or brand-short titles only. MUST NOT edit files, create brand strategy, check trademark availability, write long marketing copy, or rename code identifiers. For those adjacent requests, MUST stop title generation and route using the "Do Not Use" mapping.

Do Not Use

Route adjacent requests before generating titles:

• Slogans, taglines, ad copy, landing-page copy, or long marketing copy -> use a copywriting workflow.
• SEO page titles, keyword headlines, or search-snippet optimization -> use an SEO/content workflow.
• Trademark availability, legal clearance, naming conflicts, or registration advice -> tell the user this needs legal review.
• Code identifiers, variable names, package names, class names, or refactor naming -> use a code naming/refactor workflow.
• Full brand strategy, positioning, naming architecture, tone system, or brand book work -> use a brand strategy workflow.

Execution

CREATE A TODO LIST FOR THE TASKS BELOW. Keep the list internal unless the user asks to see process.

1. Read the user's input. If no product, feature, concept, or direction is provided, ask one concise question for the missing target and stop.
2. If the input matches any "Do Not Use" case, state that this skill only generates short product titles, route using that mapping, and stop.
3. If the user's title or naming request is ambiguous between a short product title, SEO headline, slogan, campaign copy, full product name, or brand strategy, ask one concise clarification question and stop.
4. If the input contains multiple unrelated products, split them into separate targets and use one "Multi-Product Output Format" block per clear target; if any target is unclear, ask the user to choose the target and stop.
5. For each clear target, normalize the input into four fields: product object, core capability, use scenario, and desired tone. If a field is missing, infer it from the provided text without inventing unrelated positioning.
6. If the product object exists but core capability and use scenario cannot be inferred from the input without inventing unrelated positioning, ask one concise question for the missing capability or scenario and stop.
7. Route the title style for each target. If product object, source brand, scenario, and core capability imply different routes, prioritize core capability first, then use scenario and desired tone to refine wording:

• Learning products -> companionship, sprint, rescue, training, improvement.
• Technical platforms -> base, platform, hub, engine, cockpit, infrastructure.
• Data or observability products -> observation, insight, monitoring, tracing, visibility.
• Launch or incubation products -> launch, incubation, startup, publishing, product desk.
• Code or development products -> repository, code, understanding, navigation, insight.
• Otherwise -> use a neutral product-entry style.

1. Generate and refine candidates until the final visible set contains exactly one recommendation and six unique alternatives, unless the user explicitly requests a different count. Each candidate MUST preserve the core object or core capability.
2. Filter candidates with the title rules below. Remove titles that are too long, too generic, too marketing-heavy, awkwardly translated, or semantically off-target. If too few valid titles remain for one recommendation plus the required number of alternatives, generate more candidates and repeat filtering until the output can be filled. If the user's explicit constraints make the required count impossible, ask one concise clarification question or state the conflict and stop.
3. Select the recommendation using this priority order: semantic fit, product-entry feel, compactness, distinctiveness, and natural Chinese phrasing.
4. Output the exact format in "Output Format" or "Multi-Product Output Format" and end, unless the user explicitly asks for analysis, more options, fewer options, or a different format.

Failure exits and overrides:

| Condition | Handling | Output shape | | --- | --- | --- | | Empty or missing target | Ask for the product, feature, module, or concept and stop. | 请告诉我需要命名的产品、功能、模块或概念。 | | Any request matching "Do Not Use" | State that this skill only generates short product titles, route using the mapping, and stop. | 这个 skill 只生成短产品标题。这个请求属于 ，请使用 。 | | Ambiguous title or naming intent | Ask the user to choose the intended output type and stop. | 你想要的是产品入口短标题、SEO 标题、slogan/营销文案，还是品牌命名？ | | Multiple unrelated products but not each product is clear | Ask the user to choose the target product or split the request and stop. | 我看到了多个可能的命名对象，请先指定要命名哪一个产品、功能或模块。 | | Conflict between user-requested length and this skill's 4-6 unit rule | Follow the user's explicit length requirement and mention it in the reason. | Use the normal output format, with the length override mentioned in the recommendation reason. |

Title Rules

• Unless the user explicitly requests another length, each title MUST be compact: 4-6 Chinese characters or 4-6 display units.
• Count display units this way: each Chinese character counts as one unit; each contiguous English word, acronym, brand token, or meaningful number counts as one unit; spaces, hyphens, and punctuation do not count.
• Count AI, Agent, GitHub, Lab, and meaningful numbers as one display unit each when they are necessary to preserve the source meaning. For example, AI观测台, Agent观测台, and GitHub洞察台 are each four units.
• Prefer Chinese titles. Keep English tokens only when they are the product object, established industry wording, or explicitly requested by the user.
• The recommendation MUST NOT appear again in the alternatives.
• The alternatives list MUST contain exactly six titles unless the user explicitly requests a different number of alternatives.
• All titles MUST feel like real product, module, or navigation-entry names, not explanatory phrases.
• All titles MUST keep the source direction. Do not make a title sound better by changing the product category or core capability.
• A title is too generic when it does not preserve the source product object, product category, or core capability.
• A title is too marketing-heavy when it depends on hype words such as 神器, 王者, 爆款, 超级, or 未来 instead of the product object or capability.
• A title is awkwardly translated when it follows source-language word order in a way a native Chinese UI title would not use.
• A title is semantically off-target when it changes the product category, drops the core capability, or replaces the requested scenario with a different one.

Style Patterns

Use these structures first:

• Noun + noun: 代码驾驶舱, 智能底座
• Verb + noun: 读懂仓库, 洞察团队
• Adjective + noun: 智能观测层, 极速陪练
• Scenario + capability: 仓库洞察, 日语冲刺

Avoid

WRONG: 未来之眼, 超级平台 Reason: too abstract; the user cannot infer product function. GOOD: 仓库洞察, 智能底座

WRONG: 爆款神器, 效率王者 Reason: marketing-heavy and not suitable for product UI. GOOD: 智能发布台, 考前陪练

WRONG: 智能基础设施给 AI Agent Reason: awkward translation and too long. GOOD: 智能基建, 智能AI底座

WRONG: 团队异构智能体可观测平台 Reason: explanatory phrase, not a compact product title. GOOD: 智能观测层, 团队观测台

WRONG: turning "observability layer" into 协作平台 Reason: semantic drift from observability to collaboration. GOOD: 观测中枢, 异构观测

Output Format

推荐标题：XXXX

备选标题：
1. XXXX
2. XXXX
3. XXXX
4. XXXX
5. XXXX
6. XXXX

推荐理由：
一句话说明为什么推荐该标题，重点解释它如何贴合产品定位、核心能力或使用场景。

Output only this block unless the user explicitly asks for analysis, more options, fewer options, or a different format.

Multi-Product Output Format

When the input contains multiple clear source concepts, output one repeated block per source concept using this exact format:

## 

推荐标题：XXXX

备选标题：
1. XXXX
2. XXXX
3. XXXX
4. XXXX
5. XXXX
6. XXXX

推荐理由：
一句话说明为什么推荐该标题，重点解释它如何贴合该输入概念的定位、核心能力或使用场景。

Output only these repeated blocks unless the user explicitly asks for analysis, more options, fewer options, or a different format.

Examples

Example 1

Input:

GitHub Learning Lab · 代码库理解驾驶舱

Output:

推荐标题：代码驾驶舱

备选标题：
1. 仓库洞察
2. 读懂仓库
3. 代码领航台
4. 项目透视
5. 仓库导航
6. 代码洞察

推荐理由：
代码驾驶舱最贴合代码库理解与全局掌控的产品定位，既有开发场景感，也有平台化入口感。

Example 2

Input:

深大日语 · 30 日抢救陪练

Output:

推荐标题：日语抢救营

备选标题：
1. 30日陪练
2. 日语冲刺
3. 考前陪练
4. 日语急救课
5. 深大日语练
6. 冲刺陪练

推荐理由：
日语抢救营更有紧迫感和陪伴感，适合短周期、强目标的学习陪练产品。

Example 3

Input:

Intelligence Infrastructure for AI Agent

Output:

推荐标题：智能基建

备选标题：
1. AI基础台
2. 智能底座
3. 智能AI底座
4. 智能中枢
5. 智能引擎
6. Agent智能底座

推荐理由：
智能基建准确承接 Intelligence Infrastructure 的含义，简洁、有平台感，适合作为 AI Agent 底层能力标题。

Example 4

Input:

AI 产品发布平台 + Agent 平台

Output:

推荐标题：智能发布台

备选标题：
1. AI发布台
2. Agent产品台
3. 产品孵化台
4. 智能产品台
5. 发布智能体
6. 产品启动台

推荐理由：
智能发布台同时覆盖 AI 产品发布和 Agent 平台能力，适合作为产品入口或平台模块标题。

Example 5

Input:

团队异构 Agent 的可观测性层

Output:

推荐标题：智能观测层

备选标题：
1. Agent观测台
2. 团队观测台
3. 异构观测层
4. 协同观测
5. 智能监控台
6. 观测中枢

推荐理由：
智能观测层保留了可观测性的技术含义，同时具备更自然的产品化表达。

Example 6

Input:

给“团队异构 Agent 的可观测性层”起一个 8 个字的产品标题

Output:

推荐标题：团队智能观测中枢

备选标题：
1. 异构智能观测平台
2. 团队智能监控平台
3. 多元智能观测中枢
4. 异构团队观测平台
5. 智能协同观测中枢
6. 多源智能观测平台

推荐理由：
团队智能观测中枢保留了团队、异构 Agent 与可观测性的核心方向；本次按用户明确要求采用 8 个字标题。

Example 7

Input:

帮我给这个变量起名：userAuthTokenCache

Output:

这个 skill 只生成短产品标题。这个请求属于代码标识符命名，请使用 code naming/refactor workflow。

Example 8

Input:

GitHub Learning Lab · 代码库理解驾驶舱
团队异构 Agent 的可观测性层

Output:

## GitHub Learning Lab · 代码库理解驾驶舱

推荐标题：代码驾驶舱

备选标题：
1. 仓库洞察
2. 读懂仓库
3. 代码领航台
4. 项目透视
5. 仓库导航
6. 代码洞察

推荐理由：
代码驾驶舱最贴合代码库理解与全局掌控的产品定位，既有开发场景感，也有平台化入口感。

## 团队异构 Agent 的可观测性层

推荐标题：智能观测层

备选标题：
1. Agent观测台
2. 团队观测台
3. 异构观测层
4. 协同观测
5. 智能监控台
6. 观测中枢

推荐理由：
智能观测层保留了可观测性的技术含义，同时具备更自然的产品化表达。

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: tranfu-labs
• Source: tranfu-labs/tranfu-skills
• License: MIT

Install and usage instructions live in the source repository linked above.