What problem does it solve? Research knowledge from literature reviews, ideation, and experiments is usually scattered across sessions and forgotten, forcing repeated re-derivation. This Skill builds a persistent, per-project knowledge base that compounds papers, ideas, experiments, and claims into a queryable graph across the entire research lifecycle. ## Core Features & Use Cases - Structured entity management: Stores papers, ideas, experiments, and claims as typed nodes with canonical IDs, plus a typed relationship graph (extends, contradicts, supports, invalidates) in edges.jsonl. - Subcommand workflow: Provides init, ingest, sync, query, update, lint, and stats subcommands backed by a single canonical helper script (research_wiki.py) for deterministic operations. - Compressed query pack: Generates a hard-budgeted 8000-character query_pack.md summarizing gaps, failed ideas, top papers, and open questions for downstream ideation. - Use Case: After reading a batch of arXiv papers, run the ingest subcommand to create deduplicated paper pages with metadata from the arXiv API, wire relationship edges, and later query the wiki to seed new idea generation while avoiding previously failed directions. ## Quick Start Ask the assistant to initialize the research wiki with "/research-wiki init" and then ingest an arXiv paper by ID to create its first paper page.
