paper-note-review

name	paper-note-review
description	Recover section-level evidence from existing ResearchOS paper notes for one claim or manuscript section. Use when a draft, idea, or review needs grounded evidence without re-reading the entire literature corpus or inventing citations.
tools	["read_file","write_file","list_files","grep_search","build_section_evidence_supplement","finish_task"]
strict_tools	true
model_tier	standard
temperature	0.1
allowed_read_prefixes	["user_inputs/paper-note-review/","literature/","drafts/","_runtime/skill_sessions/"]
allowed_write_prefixes	["literature/","drafts/"]
outputs_expected	{"evidence_report":"literature/skill_note_evidence_report.md","evidence_record":"literature/skill_note_evidence_report.json"}
interaction	{"mode":"guided","language":"zh-CN","summary":"围绕论文中的一个具体位置或 claim 回查已有论文阅读笔记，并给出可引用、仅作背景或无法支持的明确结论。","request_required":true,"request_prompt":"请说明要写/审查的 section 或 claim、语言，以及需要支持、质疑还是限定该主张。","example_request":"为英文 Introduction 中“agent memory 会造成跨任务 carryover”的机制表述回查现有笔记，列出可用 citation、边界和不足证据。","required_inputs":[{"id":"evidence_question","label":"Claim 或章节问题","description":"写明目标 section、待支持/质疑的主张、允许的措辞强度和任何已知 citation key。","paths":"[Truncated]","extensions":"[Truncated]","min_bytes":40,"example":"user_inputs/paper-note-review/question.md"}],"optional_inputs":[{"id":"resource_index","label":"论文资源索引","description":"可选；有该文件时先生成 section evidence supplement，避免无目标扫描笔记目录。","paths":"[Truncated]","extensions":"[Truncated]","min_bytes":80,"example":"drafts/manuscript_resource_index.json"}],"outputs":[{"id":"evidence_report","label":"Section 证据报告","path":"literature/skill_note_evidence_report.md","description":"精确 note section、可用/不可用证据、建议措辞和升级路径的可读报告。"},{"id":"evidence_record","label":"结构化证据记录","path":"literature/skill_note_evidence_report.json","description":"每个候选 source 的 note path、heading、证据等级、允许用法和 claim 结论。"}]}

Section-Level Note Evidence Review

Read the evidence question. If a manuscript resource index exists, call build_section_evidence_supplement for the closest section first. Use the supplement/card path to run grep_search for the named headings, then read_file only for the relevant note section. Record exact paths and headings.

Classify each source as claim_usable, background_or_boundary_only, or insufficient. Never turn abstract-only or weak cards into mechanism proof. If exact sections do not support the requested wording, recommend a weaker wording, removal, or a bounded query plan with an explicit evidence gap; do not fabricate a citation.