--- name: autorebuttal description: Automatically plan, draft, and compliance-check a complete evidence-grounded rebuttal from a confirmed baseline. --- # AutoRebuttal Read these shared references completely before acting: - `/app/skills/autorebuttal-shared/references/input-contract.md` - `/app/skills/autorebuttal-shared/references/strategy-and-budget.md` - `/app/skills/autorebuttal-shared/references/drafting-and-compliance.md` - `/app/skills/autorebuttal-shared/references/evidence-policy.md` - `/app/skills/autorebuttal-shared/references/artifact-contracts.md` Read the confirmed intake and analysis from the rebuttal context before doing any generation. This author-facing phase contains three internal steps that must run in order without asking the author to type “continue.” ## Step 1: Response strategy Resolve shared issues, priority reviewers, response moves, output structure, and a reviewer-complete character budget. The format and limits must match the confirmed baseline. Keep a safety margin and never invent venue policy. ## Step 2: Drafting Write concise reviewer-addressed entries covering every analyzed concern exactly once across the entry set. Ground factual claims in the paper or explicit author evidence. Use visible `[AUTHOR: ...]` placeholders when only the author can provide a result or decision. ## Step 3: Compliance review Audit and refine the full draft for concern coverage, reviewer consistency, evidence integrity, response-character limits, and submission readiness. Fix non-blocking writing problems directly. Keep unresolved evidence visible and set `ready_for_submission` false when author action remains. Publish the complete AutoRebuttal phase contract exactly once with `publish_stage_result` and idempotency key `autorebuttal-v1`.