survey-evidence-package

name	survey-evidence-package
description	Prepare a human-approved, evidence-audited survey package from a domain synthesis before any Survey prose is drafted. It checks corpus sufficiency, proposes taxonomy and storyline, identifies weak classes, and can request or perform scoped literature supplementation. Use when a researcher wants a safe bridge from T3.5-style synthesis to T3.6 survey writing.
tools	["read_file","write_file","write_structured_file","list_files","grep_search","multi_source_search","openalex_search","semantic_scholar_search","arxiv_search","crossref_search","build_survey_state","expand_corpus_for_survey","ask_human","update_skill_workflow","finish_task"]
strict_tools	true
model_tier	standard
temperature	0.15
allowed_read_prefixes	["project.yaml","user_inputs/survey-evidence-package/","literature/","drafts/survey/","user_seeds/","_runtime/skill_sessions/"]
allowed_write_prefixes	["user_inputs/survey-evidence-package/","literature/","drafts/survey/"]
outputs_expected	{"corpus_sufficiency":"drafts/survey/corpus_sufficiency_report.md","taxonomy_candidates":"drafts/survey/taxonomy_candidates.json","storyline":"drafts/survey/survey_storyline.md","survey_package":"drafts/survey/survey_evidence_package.json","supplementation_plan":"drafts/survey/survey_supplementation_plan.md","workflow_manifest":"drafts/survey/survey_evidence_package_manifest.json"}
interaction	{"mode":"guided","language":"zh-CN","summary":"在写 Survey 前审计语料充分性、taxonomy、故事线和弱证据；用户可选择先补检、仅保守写作或确认进入 T3.6。","request_required":true,"request_prompt":"请说明 Survey 的中心问题、目标读者/venue、希望的语言与范围，以及是否允许系统针对薄弱 taxonomy 类做定向检索。","example_request":"基于现有领域综合准备英文 Survey；先告诉我哪些 taxonomy 类证据不够，并允许我确认后补检。","required_inputs":[{"id":"survey_intent","label":"Survey 意图与范围","description":"说明中心问题、读者、边界、语言/模板偏好和是否准备现在进入 Survey 写作。","paths":"[Truncated]","extensions":"[Truncated]","min_bytes":60,"example":"user_inputs/survey-evidence-package/intent.md"}],"optional_inputs":[{"id":"synthesis_sources","label":"领域综合与文献证据","description":"可选；优先使用现有 synthesis/workbench/note cards。若没有，系统会要求先运行领域综合或上传可审计来源。","paths":"[Truncated]","extensions":"[Truncated]","min_bytes":60,"example":"literature/domain_synthesis.json"}],"outputs":[{"id":"corpus_sufficiency","label":"Survey 语料充分性报告","path":"drafts/survey/corpus_sufficiency_report.md","description":"FULL/PARTIAL/ABSTRACT/METADATA 覆盖、taxonomy 类缺口、不可主张边界和推荐动作。"},{"id":"taxonomy_candidates","label":"Taxonomy 候选","path":"drafts/survey/taxonomy_candidates.json","description":"可审计 taxonomy 选项、分类维度、paper links、覆盖程度与冲突。"},{"id":"storyline","label":"Survey 叙事与章节主张","path":"drafts/survey/survey_storyline.md","description":"问题、rationale、taxonomy 演进、每节问题和证据限制；不是最终 TeX。"},{"id":"survey_package","label":"Survey 证据包","path":"drafts/survey/survey_evidence_package.json","description":"后续 T3.6 计划可读取的证据索引、taxonomy 决策、弱证据和已确认约束。"},{"id":"supplementation_plan","label":"定向补检/补读计划","path":"drafts/survey/survey_supplementation_plan.md","description":"哪些类需要何种来源、为什么、可查询什么以及补齐后允许升级的结论。"},{"id":"workflow_manifest","label":"Survey 证据工作流清单","path":"drafts/survey/survey_evidence_package_manifest.json","description":"输入、查询尝试、证据限制、人工决策和 T3.6 handoff 状态。"}]}
workflow	{"kind":"integrated","summary":"在写作前先审计综述的证据与 taxonomy；允许定向补检，但不把补检线索直接升级为 Survey 论证。","phases":[{"id":"survey_contract","label":"Survey 意图与证据盘点","objective":"明确综述问题、目标读者、语言和现有综合产物。","operations":"[Truncated]"},{"id":"sufficiency_review","label":"语料充分性与 taxonomy 候选","objective":"判断是否能形成可解释 taxonomy，识别薄弱或空白类。","operations":"[Truncated]"},{"id":"supplementation_gate","label":"补检/保守范围决策","objective":"让用户选择只用当前语料、授权定向补检、上传材料或停止。","operations":"[Truncated]","human_gate":true},{"id":"survey_handoff","label":"Survey 写作交接","objective":"输出可审计 evidence package，并仅在用户确认后标记可进入 T3.6。","operations":"[Truncated]","human_gate":true}]}

Survey Evidence Package

Do not write a survey manuscript in this Skill. Treat it as the pre-writing evidence gate that a researcher can inspect. Create a corpus sufficiency report before choosing a taxonomy. It must distinguish unavailable retrieval, metadata-only lead, abstract-only coverage, and full/partial evidence suitable for bounded discussion.

When proposing the Survey storyline and taxonomy, prepare the conditions for clean publication prose rather than a long catalogue of papers. The eventual survey should be able to move naturally from the review question and the field's consequential developmental shifts into the taxonomy, comparisons, challenges, and future agenda. Keep the planned section hierarchy compact, and do not reserve a separate heading for each paper or taxonomy label. Each proposed section needs a reader question, an argument, a bridge from the prior section, and a note of where terms need first-use explanation or a real/conditional such as example. The later writer should use complete, connected sentences rather than dash/colon label prose or fragmented lists. This is a writing preference, not a mechanical pass/fail rule.

For every proposed source, preserve the evidence level and the specific claim types it can support. Recommend broad, high-quality coverage from genuinely relevant UTD/FT50/ CCF-A, leading disciplinary, foundational, highly cited, influential, and important recent work, but never pad a taxonomy with prestigious papers that do not support its claims. Citation existence, source meaning, and claim-level fit must be verified again by the T3.6 writer before the source enters Survey prose.

If the human authorizes supplementation, make small source-returning searches and record the exact query, retained source, and what it can upgrade. Do not convert search results into central taxonomy evidence without a readable source/card. The final handoff must state whether it is ready for conservative Survey planning, ready only after targeted reading, or not ready.