菜单
Classify search results by what type of product decision they can support, so the agent does not upgrade weak sources into hard facts.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Diagnose whether comments contain demand signals for explanation, comparison, tutorial, retest, or follow-up content.
Synthesize search results into source-backed facts, agent inference, and evidence gaps.
Diagnose whether an old creator post contains reusable ReCast asset value and what kind of asset it is.
Judge whether the old content deserves a ReCast task, no-action report, or further observation.
Frame the likely ReCast opportunity type before external research: retest, comparison, tutorial update, recommendation update, follow-up, or no action.
Compose a natural-language research plan from the old content diagnosis, opportunity hypothesis, and candidate probe skills.
| name | source_quality_policy |
| description | Classify search results by what type of product decision they can support, so the agent does not upgrade weak sources into hard facts. |
| stage | evidence |
This skill exists to protect product judgement. ReCast should not treat all search results as equal.
Use for claims that require high confidence:
This tier normally requires an official product/company/documentation domain, or a domain explicitly constrained by the research plan.
Use for:
This tier can justify why a topic is worth revisiting, but cannot alone prove official product facts.
Use for:
This tier can prove that people are discussing something. It cannot prove that a model version, policy, or product update is real.
Use for weak or ambiguous sources:
Do not use this tier as factual support. Mention as a gap or weak signal only.
Use for sources that should be excluded from factual reasoning:
Before writing a finding, ask:
The final task card can use weak sources to motivate investigation, but it must not present weak sources as established truth.
