Name: semantic_selector
Author: richardlee1106

Imported from richardlee1106/geoloom-agent (backend/SKILLS/SemanticSelector/SKILL.md). Install upstream with npx skills add richardlee1106/geoloom-agent --skill SemanticSelector. Copyright stays with the author.

SemanticSelector Skill

Prompt

area insight 按需取证技能。用于在已经拿到结构化区域证据后，按当前 query 的语义焦点选择真正需要的类别、竞争层和代表样本。它不是黑名单过滤器，也不是 prompt 补丁；核心职责是让模型在“业态结构 / 咖啡店 / 公共厕所分布 / 某类配套”这类问题里，只保留当前问题真正相关的证据。优先在这些场景调用：

用户问题只关心某一类主题，而不是整片区域的全量总结
已拿到 area insight，但当前证据太杂，容易把回答带偏
你需要把“按需取证”交给语义选择而不是手写规则调用前优先先拿到 area_category_histogram、area_representative_sample，必要时再带上 competition_density / AOI / landuse。

{ "ocm": "1", "id": "richardlee1106-geoloom-agent-semanticselector", "kind": "skill", "name": "semantic_selector", "description": "area insight 按需取证技能", "publisher": "richardlee1106", "version": "1.0.0", "capabilities": { "domains": [ "data_analysis" ], "tags": [ "skill-md", "github" ], "languages": [ "en" ] }, "quality_prior": 0.6, "examples": [ "area insight 按需取证技能" ], "primary": false, "metadata": { "source": { "provider": "github", "repository": "https://github.com/richardlee1106/geoloom-agent", "path": "backend/SKILLS/SemanticSelector/SKILL.md", "ref": "f018caf7c3860a75bcd990b3827525025152f37d", "url": "https://github.com/richardlee1106/geoloom-agent/blob/f018caf7c3860a75bcd990b3827525025152f37d/backend/SKILLS/SemanticSelector/SKILL.md", "key": "richardlee1106/geoloom-agent/backend/SKILLS/SemanticSelector/SKILL.md" } }, "instructions": "# SemanticSelector Skill\n\n## Prompt\narea insight 按需取证技能。用于在已经拿到结构化区域证据后，按当前 query 的语义焦点选择真正需要的类别、竞争层和代表样本。\n它不是黑名单过滤器，也不是 prompt 补丁；核心职责是让模型在“业态结构 / 咖啡店 / 公共厕所分布 / 某类配套”这类问题里，只保留当前问题真正相关的证据。\n优先在这些场景调用：\n- 用户问题只关心某一类主题，而不是整片区域的全量总结\n- 已拿到 area insight，但当前证据太杂，容易把回答带偏\n- 你需要把“按需取证”交给语义选择而不是手写规则\n调用前优先先拿到 area_category_histogram、area_representative_sample，必要时再带上 competition_density / AOI / landuse。", "cost": { "context_tokens": 98 } }

semantic_selector

About

SemanticSelector Skill

Prompt

Use it

Manifest

Reviews