research-analysis

Verify scientific drafts against their source material. Checks every factual claim, adds or corrects inline citations, marks unverified or tentative statements, and removes unsupported numerics. Run after the writer agent on papers/<slug>.md before delivery.

Draft scientific papers and research documents from collected notes and source material. Use when research has already been gathered and needs to be turned into a structured manuscript. Produces a polished draft in papers/<slug>.md following IMRAD structure with full paragraphs, inline citations, and a Sources appendix.

Search and retrieve arXiv academic papers by topic, category, or paper ID — with AlphaXiv-enriched AI-generated overviews. Uses arXiv Atom API across cs.AI/cs.LG/cs.CL/cs.CR/cs.MA/cs.SE/cs.IR. Three workflows: Latest, Search, Paper. USE WHEN arxiv, papers, latest papers, research papers, recent ML papers, paper lookup, summarize paper, latest LLM papers, AI safety papers, cs.AI latest. NOT FOR general research (Research), URL parsing (_PARSER), or annual reports (_ANNUALREPORTS).

Multi-agent collaborative debate producing visible round-by-round transcripts with real intellectual friction. Members custom-composed via ComposeAgent with domain expertise tailored to the topic — never generic. Workflows: DEBATE (3 rounds, full transcript + synthesis, parallel within rounds, 40-90s), QUICK (1 round, fast perspective check). 4-6 well-composed agents outperform 12 generic ones. Collaborative-adversarial — debate to find the best path. USE WHEN council, debate, multiple perspectives, weigh options, deliberate, get different views, what would experts say, pros and cons. NOT FOR pure adversarial attack (use RedTeam).

Content-adaptive wisdom extraction — reads content first, detects what wisdom domains are present, then builds custom sections around what it finds instead of forcing static headers. A security talk gets 'Threat Model Insights'; a business podcast gets 'Contrarian Business Takes'. Five depth levels: Instant, Fast, Basic, Full (default 5-12 sections), Comprehensive (10-15+ themes). Output always includes dynamic sections, One-Sentence Takeaway, 'If You Only Have 2 Minutes', References. Spicy/contrarian takes mandatory. YouTube via fabric -y; articles via WebFetch. Workflow: Extract. USE WHEN extract wisdom, analyze video, analyze podcast, extract insights, key takeaways, summarize interview, distill content. NOT FOR static Fabric extract_wisdom pattern (use Fabric).

Execute any of 240+ specialized prompt patterns natively across Extraction, Summarization, Analysis, Creation, Improvement, Security, Rating. Common: extract_wisdom, create_threat_model, analyze_claims, improve_writing, review_code, mermaid, youtube_summary. CLI used only for YouTube transcript (-y) and URL fallback (-u). Two workflows: ExecutePattern, UpdatePatterns. USE WHEN fabric, fabric pattern, run fabric, update patterns, threat model, analyze claims, improve writing, review code, mermaid, STRIDE, sigma rules. NOT FOR multi-agent investigation (Research) or content-adaptive extraction (ExtractWisdom).

Physics-based reasoning framework (Musk methodology) that deconstructs problems to irreducible fundamental truths rather than reasoning by analogy. Three steps: DECONSTRUCT (break to constituent parts and actual values), CHALLENGE (classify every element as hard constraint / soft constraint / unvalidated assumption — only physics is truly immutable), RECONSTRUCT (build optimal solution from fundamentals alone, ignoring inherited form). Outputs: parts breakdown, constraint table, reconstructed solution. Workflows: Deconstruct, Challenge, Reconstruct. USE WHEN first principles, fundamental truths, challenge assumptions, real constraint, rebuild from scratch, start over, physics first, question everything, reasoning by analogy. NOT FOR structural feedback loops (use SystemsThinking).