research-surveylisted

# Research Survey Pipeline Orchestrate a full literature survey: vec-db semantic search → web search → download → parallel subagent reading → structured report. ## Overview This skill automates the research paper discovery-and-analysis workflow. It combines a local vector database of ~60K+ indexed top-venue papers with web search for broader coverage, then deploys **parallel subagents** to deep-read each paper and produce a structured comparative analysis anchored to the user's research idea. The goal is not just to list papers, but to produce actionable competitive intelligence: how does each paper relate to the user's proposed work, what can be borrowed, what differentiates the user's approach. ## Prerequisites - **Vec-db**: LanceDB semantic search database at `/home/vla-reasoning/proj/litian-research/vec-db/` - Query: `cd /home/vla-reasoning/proj/litian-research/vec-db && npx tsx src/cli.ts search "<query>" --top <N>` - Status: `cd /home/vla-reasoning/proj/litian-research/vec-db && npx tsx src/cli.ts status` - **arXiv access**: For downloading PDFs (`wget https://arxiv.org/pdf/<id> -O <id>.pdf`) - **Subagents**: Agent tool — **MUST launch all reading agents in ONE message for true parallelism** - **Target note**: A markdown file to update with findings (user specifies path) ## Workflow ### Step 1: Understand the Research Context Before searching, gather from the conversation or by asking: 1. **Research topic/idea** — Read the user's idea note if it exists 2. *