← ClaudeAtlas

researcherlisted

Runs multi-source investigations using a structured evidence-mapping cycle. Routes between business research, academic research, and claim verification. Triggers on: "research", "investigate", "verify", "compare", "analyze", "look into", "dig into", "fact check", "due diligence", "deep dive", "what is really happening", "how do real practitioners see this", "is this true", "is this legit", "what's the real story", "market landscape", "state of the art", "literature review", "competitive analysis", "opportunity scan", "risk assessment", "研究", "调研", "调查", "分析", "验证", "核实", "深挖", "摸底", "行业分析", "市场调研", "竞品分析", "尽调", "尽职调查", "这个靠谱吗", "真实情况是什么", "帮我查查", "帮我看看", "有没有前景", "值不值得", "到底怎么样", "是真的吗", "论文调研", "文献综述", "技术选型", "机会分析".
droftgamer-wq/researcher-skill · ★ 0 · AI & Automation · score 75
Install: claude install-skill droftgamer-wq/researcher-skill
# Research Operating System Investigate the internet as a mixed evidence environment, not as a pile of search results. Every task follows the same core cycle; domain-specific tactics come from reference packs loaded on demand. ## Core Cycle ```text FRAME → MAP → FRONTIER → DEEPEN → CHALLENGE → SYNTHESIZE ``` ### 1. FRAME Define the task before searching: - Decision question or judgment target - Key claims that may need confirmation or falsification - Important stakeholders and their incentives - Source classes likely to matter - What would change the answer **Load reference packs by task type:** | Task signal | Load | |-------------|------| | Industry, company, career, opportunity, operator reality | [references/business.md] | | Papers, benchmarks, state of the art, technical lineage | [references/academic.md] | | "Is this true?", fact check, verify claim, debunk narrative | [references/claim-verification.md] | | Mixed task | Load multiple packs | When the task is a **decision between options** (choose A vs B), use the core cycle but frame everything around decision criteria — see the Decision Support output variant at the end of this file. ### 2. MAP Run a broad first pass across multiple evidence classes using **WebSearch**. Aim to surface: - Official narrative - Behavioral reality (what actors do, not just say) - Operator or expert view - Lived experience - Adversarial or contradiction path **Budget: 4-6 searches.** Stop and write a field map before going dee