What problem does it solve? Academic papers often fail because prose gets written before the underlying argument is settled, forcing large rewrites when the structure moves. This Skill treats the outline as a structured database of claims, warrants, counterpoints, and boundaries, and audits it with a deterministic checker so argument holes are found before any prose exists. ## Core Features & Use Cases - Argument-first outlining: Separates clerical fields (filled by the agent from chat) from argument fields (claim, warrant, counterpoint, boundary) that only the author can supply, withholding the agent's own draft until the author answers. - Deterministic validation: Runs research-core validate outline.json --kind outline to check containment, fact/opinion tagging, claim strength against a named dictionary, hanging references, extraction provenance, and arrangement, with a documented waiver mechanism. - Reconciliation logging: Records each author-versus-agent reconciliation to a JSONL log and recomputes graduation state every run without ever lowering check severity. - Use Case: A researcher drafting a results section uses the Skill to iteratively answer one question at a time about missing warrants and counterpoints, then runs the checker to confirm every promise is discharged before writing prose. ## Quick Start Ask the agent to start a paper outline audit with the develop-paper-outline skill and run the checker on your outline.json file.
