Research-Pilot

Most research ideas die slowly. Weeks go into reading before the question is judged worth pursuing, and the choice of journal is often an afterthought.

The command is /position-idea. It sets up a folder for the idea and runs three agents in sequence. A literature scout searches Crossref and Google Scholar and writes a brief. An idea evaluator checks the idea against that brief and a fixed rubric, plus a scoop check, and gives a verdict: pursue, reframe, park or drop. A human gate follows the verdict. A journal positioner then proposes a stretch target, a realistic target and at least one ABDC A journal, with a positioning of the question and five to ten must-read papers. The main thread writes everything into a plan file.

It runs on Claude.

One idea folder, dated 23 October 2026, has been through the full pipeline. The verdict it produced is the output; whether a journal agrees is a slower test than this system can run.

Putting the verdict before the journal choice is deliberate: a polished positioning of a weak idea is wasted effort. The weakness is that the scout reads abstracts and snippets, so its brief can only be as good as what it can reach. I treat the verdict as a prompt to think, not a decision.