Moonweave-Research

Prevents citation hallucination in academic writing. Invoke when: finding papers to support a specific claim; verifying/checking/auditing existing citations or DOIs; confirming whether a paper actually says what the user claims it says ('is that what the paper says?', 'did they actually show X?'); adding a citation by describing a paper ('add a citation for the paper where X'); running a pre-submission reference sweep. Do NOT invoke for: formatting references in APA/IEEE style, general topic explanations, citation style questions, or prose editing. Selects Quick Screen (seconds per paper) or Full Audit (abstract fetch + claim check) automatically.

Proactively detects contradictions between claims across a set of papers. Invoke when the user wants to find conflicts in a research area, asks 'do these papers agree?', 'what's the debate around X?', 'find contradictions in my literature set', provides a list of DOIs or a search query for conflict analysis, or is writing a literature review and wants to identify contested claims. Do NOT invoke for summarizing papers, finding papers on a topic (use arxiv-monitor), or verifying a single citation (use ref-verify).

Verifies whether a scientific figure actually supports quantitative claims in the paper text. Invoke when the user asks 'does Figure X show Y?', 'the paper claims Z% — does the figure actually show that?', 'is this graph consistent with the abstract?', 'check this figure against the results', or provides a figure image with a specific claim to verify. Also catches wrong figure pointers (value is correct but cited in wrong figure number) and characterizes whether a figure is quantitative or photographic before attempting to read it. Do NOT invoke for creating figures, plotting data, or general image description.

한국연구재단(NRF) 연구계획서 작성 전용 스킬. 연구계획서 섹션 초안 작성, 심사기준(우수성·필요성·활용방안) 정렬 검토, IRIS 제출 형식 준수, 한국어 학술 문체 적용. 다음 상황에서 invoke: '연구계획서 써줘', 'NRF 신청서 작성', '필요성 섹션 검토해줘', '심사위원 입장에서 평가해줘', '연구내용 한 페이지로 정리', 특정 NRF 과제 유형 언급(신진연구, 기초연구, 선도연구센터 등). 일반 논문 writing이나 해외 그랜트(NSF/NIH)에는 invoke하지 말 것.

Fetches new arXiv papers matching a research watchlist, scores relevance with LLM, and delivers a filtered daily digest. Invoke when the user wants to check new arXiv papers on a topic, set up a daily paper monitoring routine, asks 'what's new on arXiv about X?', or wants to scan recent preprints in their field. Also use as a Claude Code Routine template for scheduled daily briefings. Do NOT invoke for finding a specific known paper (use ref-verify) or for general web research.

Prevents citation hallucination in academic writing. Invoke when: finding papers to support a specific claim; verifying/checking/auditing existing citations or DOIs; confirming whether a paper actually says what the user claims it says ('is that what the paper says?', 'did they actually show X?'); adding a citation by describing a paper ('add a citation for the paper where X'); running a pre-submission reference sweep. Do NOT invoke for: formatting references in APA/IEEE style, general topic explanations, citation style questions, or prose editing. Selects Quick Screen (seconds per paper) or Full Audit (abstract fetch + claim check) automatically.

Use for paper-figure quality, compile, export, and reproducibility gates around scientific schematics. A human or any LLM/tool may author the figure; figure-agent enforces Style Lock, compiles/exports, runs visual QA checks, and reports stale or unreplayable figure state. Deleted v0.1 prompt/image-gen commands are historical only. Symbolic schematic axes are inside scope; quantitative data plots and measured datasets belong in matplotlib / Graph_making_hub.

Use when designing, implementing, or reviewing motion for the research-wiki graph UI, especially Reels/Capture Mode, semantic graph animations, canvas node/link motion, Instagram demo shots, or visual polish that must stay meaningful and restrained. Pair with anneal when comparing multiple motion directions before implementation.