researcher-pack

The LLM-research toolkit you actually have is a bag of disconnected tools — a chatbot for brainstorming, a summarizer for papers, an autocomplete for writing. Each is individually impressive. Nothing compounds.

You park an idea in a chat log on Monday, read a paper in a PDF reader on Wednesday, draft a paragraph in a LaTeX editor on Friday, and by the following week none of these pieces know the others exist. The chat is gone. The paper notes are in some download folder. The draft has drifted.

*researcher-pack is one loop, and you only ever type one command.*

The one command

/research-session is the front door. When you type it, the skill reads .claude/research-state.yaml, scans the wiki for stale topic pages and unrevisited research evaluations, pulls unresolved review findings from recent events, and produces a briefing in under a page.

The briefing tells you three things: what's due, what drifted, what to pick up. It proposes an agenda, and from that agenda it dispatches to the right sub-skill — /paper-read if you want to ingest a single paper, /lit-search if you want to map a subfield, /research-companion if you want to brainstorm, /weekly-review if you want a digest, /orchestrate if the job is a multi-stage pipeline.