cs-run: One Front Door for Vague Agent Requests

Where this fits

A user drops a vague request: "帮我处理一下这个功能" or "I need something for the landing page." They don't name a skill, and you don't know if they want design, frontend, copy, or a plan. cs-run is the entrypoint for that moment. It turns the rough intent into a compact Goal Card, asks only the blocking questions, and routes to exactly one active downstream skill so execution can start immediately.

Why agents benefit

• Resolves ambiguity before spending tokens: it extracts a Goal Card and asks only the questions that change routing decisions.
• Prevents skill sprawl: the agent does not have to scan the whole library or guess which SKILL.md applies.
• Maintains focus: after routing, it continues through the chosen skill when the user asks to execute, avoiding context switching.
• Works across languages and vague prompts: triggers include cs-run, $cs-run, 帮我执行, 帮我处理, 不知道用哪个 skill, and general product or workflow requests.
• Reduces user load: the user states an outcome; the agent handles routing and orchestration.

Practical scenarios

Product request without a named skill

A user says "先规划再做" and describes a new onboarding flow. cs-run creates a Goal Card, asks whether the deliverable is a strategy doc or functional prototype, and routes to one planning or product skill.

Cross-functional design and engineering task

The user asks "帮我处理一下这个页面改版" without specifying a design system or implementation approach. cs-run asks one blocking question about target output, then routes to the matching design or coding skill.

Unclear entry point for a content workflow

A user shares an article draft and says "不知道用哪个 skill." The skill interprets the desired outcome, asks about tone or format if needed, and sends the task to the relevant content skill.

Add it to your agent workflow

Install the entrypoint with:

npx skills add ChenShuo2004/cs-skills --skill cs-run

Then trigger it naturally when the user intent is vague:

User: 帮我处理一下这个需求
Agent: [cs-run] Goal Card created. Blocking question: is the output a written spec or a working implementation?

Once the user answers, cs-run routes to one active skill and continues execution when the user says go.

Compare before adopting

Before adding cs-run, compare it with alternatives on a few axes:

• Quality signals: check the 118 stars and whether the author responds to issues around routing behavior.
• Maintenance freshness: review recent commits; a routing skill needs to stay in sync with the skills it dispatches to.
• Alternatives: a local convention file or a simple if/else prompt script may be enough if you only have a few skills.
• Workflow fit: does the Goal Card format match how your agent already structures tasks? If not, you may need a thin adapter.

Why it is worth tracking

cs-run is worth watching because it solves the least glamorous part of agent skill systems: deciding what to do when the user does not know what to ask for. The library is young, but the design is pragmatic and language-aware. The skill listing at https://www.openagentskill.com/skills/chenshuo2004-cs-run is a convenient place to monitor updates and community signals. Revisit it when your skill count grows past the point where manual routing becomes a bottleneck.