| name | comment_signal_diagnosis |
| description | Diagnose whether comments contain demand signals for explanation, comparison, tutorial, retest, or follow-up content. |
| stage | diagnosis |
Skill: comment_signal_diagnosis
Mission
Read comments as demand signals, not as proof.
Comments should help ReCast decide whether the audience is asking for a next content asset.
Comment Signal Types
- explanation_demand: "why is this true?"
- comparison_demand: "A or B, which is better?"
- tutorial_demand: "how do I do this?"
- skepticism: "is this actually true?"
- access_or_buying_question: "how do I get/use/buy this?"
- counterexample: "my experience differs"
- use_case_request: "can you test this scenario?"
- joke_or_noise: not actionable unless it reveals a repeated audience frame
Product Decision It Supports
This skill tells later stages whether comments can trigger follow-up content.
Rules
- A comment is a signal, not factual evidence.
- High likes or replies increase demand strength, not factual truth.
- Do not let one funny comment dominate unless it points to a real recurring question.
- Do not let one isolated low-engagement comment dominate unless it contains a concrete use case, specific counterexample, or repeated audience frame.
- Generic challenges such as "你真的会用吗" are weak skepticism by default. They become medium or high only when the engagement, replies, or concrete reasoning shows real audience demand.
- Comments with no substantive demand, such as spam, solicitation, ads, or "代充" offers, are joke_or_noise or irrelevant. They must not be classified as access_or_buying_question.
- Comment strength must combine semantic specificity, engagement, and repetition. Semantics alone is not enough to mark a signal medium.
- If a comment is low-engagement and does not ask for a concrete test, explanation, tutorial, or counterexample, keep it low or omit it from demand signals.
- Preserve the user wording when it is a strong hook.
Strength Calibration
- high: many likes/replies, repeated across comments, or asks for a specific next content task.
- medium: clear actionable demand with meaningful engagement or a concrete scenario/counterexample.
- low: isolated, generic, low-engagement, vague, joking, or weakly related signal.
Output Contribution
Contribute comment demand signals, demand strength, and possible next content questions.
