
from ai-collab-playbook210
Run a structured, no-omission review of near-final ML conference papers: section-by-section edits, prioritized P0/P1/P2 issue list, and rebuttal drafting.
Run a two-view review for ML conference submissions: (A) a section-by-section critique with concrete edits and (B) a prioritized P0/P1/P2 issue list with verification notes. It also supports drafting professional rebuttals and revision plans while enforcing guardrails such as no hallucinated citations and preservation of LaTeX semantics.
Use this skill for pre-submission QA (ICML/ICLR/NeurIPS/AAAI) when a draft is mostly complete and you want clarity, citation checks, or a camera-ready polish. Also use after receiving reviewer comments to classify feedback and draft point-by-point rebuttals. Do NOT use for initial drafting or inventing new experiments.
Best suited for agents with strong text-editing and research tooling (Claude Code, Codex, Copilot-style agents). The workflow expects access to file-reading and document inspection tools.
Pure instructional skill (no scripts) for structured ML conference paper review with section-by-section critique and P0/P1/P2 prioritized issue lists. Also supports rebuttal drafting. Well-organized with references/ modules and clear output contracts. Strong guardrails against hallucinated citations and data exfiltration. Niche but genuinely useful for ML researchers preparing submissions.
Clean skill with no security concerns. Well-structured modular architecture with progressive disclosure via references/. Two execution modes (targeted/full-parallel) is a nice design pattern. The parity guarantee approach for consolidating legacy skills is thoughtful.