What problem does it solve? Research artifacts like ideas, experiment plans, and paper drafts often need multiple rounds of critique and revision, and doing this manually loses track of what was fixed, what score each round received, and which issues remain unresolved. ## Core Features & Use Cases - Multi-round review loop: Repeatedly calls the /review skill on an artifact, parses structured feedback, applies fixes, and re-reviews until a target score is reached or convergence is detected. - Classified issue handling: Sorts actionable items into directly fixable content issues, wiki knowledge gaps, claim status updates, and out-of-scope items, updating wiki entities and graph edges where appropriate. - Improvement history report: Produces a REFINE_REPORT with the score trajectory, fixed issues table, wiki changes, unresolved issues, and suggested next steps. - Use Case: You have a paper draft in wiki/outputs/ scoring 5/10 on review. Run the refine loop with a target score of 8 to automatically apply reviewer feedback across rounds and receive a full audit trail of every change. ## Quick Start Refine the artifact at wiki/outputs/paper-draft-v1.md with a target score of 8 and a maximum of 4 rounds.
