What problem does it solve? Answering peer reviews means tracking every reviewer point, mapping it to claims and evidence, deciding what to rebut or concede, and staying inside the venue's response limit — all without losing accountability for promised changes. This Skill turns raw review files into a structured point ledger, a venue-compliant response draft, and a checklist of promises. ## Core Features & Use Cases - Point ledger generation: Parses every review in cycls/<cycle>/reviews/ (both received reviews and simulated reviews) into atomic points mapped to attacked claims and fingerprinted evidence. - Three dispositions with user control: Each point is marked rebut, promise, or concede; concessions and promises require explicit user approval, and conceded claims are downgraded to weakened in the claims ledger. - Venue-constrained drafting: Writes RESPONSE_<date>.md within the venue's response_type and response_limit, citing only fingerprinted evidence, and mirrors every promise as a checkbox in tasks/<cycle>_promises.md. - Use Case: After receiving three reviewer reports for a conference submission, run the Skill to produce a point-by-point rebuttal letter that cites existing experimental evidence, records two promised ablations as tracked checkboxes, and flags one indefensible claim as weakened. ## Quick Start Ask the assistant to draft a point-by-point response to the reviews in the current cycle, for example by saying "run stage-resp-writer for cycle 1 and draft the rebuttal".
