/respond-reviewer

This engine runs the classify-and-dispatch orchestration shape: one reviewer letter in, N labelled comments out, each routed to the specialist that should answer it, then re-evaluated by @supervisor before being assembled into a single rebuttal letter. It is the proof that OMCR can hand-off between specialists under supervisor control — Phase 2's harder counterpart to /iterate-revision's single-loop pattern.

If you are reading this because Claude Code's skill auto-discovery surfaced it: invoke it via /respond-reviewer <review-letter-path>. Do not edit the manuscript, rebuttals.json, or .claude/omcr-state/ by hand while a run is in flight.

When this skill is invoked, immediately execute the workflow below. Do not only restate or summarize these instructions back to the user.

Signature

/respond-reviewer <review-letter-path> [--manuscript <root>] [--draft-only] [--format md|latex]

Flag	Default	Purpose
`--manuscript`	`paper.json.manuscript_root`	Manuscript root the dispatched agents may edit. Falls back to `paper.json.manuscript_root`; if neither is set, phase 01 aborts.
`--draft-only`	`false`	Dispatch all per-comment responses but skip every manuscript edit. The rebuttal letter is the only artifact written.
`--format`	`latex`	Output format for the assembled rebuttal letter. `latex` → `rebuttal-letter.tex`; `md` → `rebuttal-letter.md`. Input format is auto-detected from the letter extension (Phase 2 decision §2).