Rebuttal Skill
A skill that breaks reviews of a paper into concerns, plans experiments and prepares material for a reply the author finishes.
- 507 ★ on the repository
- Last commit
Minimal access
Why this level
- The skill runs nothing and works with review texts and response drafts.
- Review and paper texts go to the chosen model.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/rebuttal-skillDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skill runs nothing and works with review texts and response drafts.
- Review and paper texts go to the chosen model.
This is third-party code. Review the repository files before installing.
What it does
Rebuttal Skill helps authors of NeurIPS, ICML, ACL and CVPR papers work through reviews with OpenCode, Claude Code or Gemini CLI. The work runs in two stages. First the skill splits reviews into atomic concerns, infers what really worries the reviewer, assigns a severity from FATAL to MINOR and builds an experiment plan with priorities P0 to P3. Second, once results are ready, it checks evidence against concerns, writes the response as "direct answer, evidence, revision" and drafts a manuscript revision list. Before the stages there is a viability check: `PROMISING`, `BORDERLINE` or `LOW EXPECTED RETURN`. For hopeless cases the skill gives a resubmission roadmap. There are five output modes, from `TRIAGE_AND_EXPERIMENT_PLAN` to `QUALITY_REVIEW` for auditing a finished draft. The skill forbids fabricating numbers or hiding negative results and marks gaps with visible placeholders.
Who it is for. Machine learning authors preparing replies to NeurIPS, ICML, ACL or CVPR reviewers who want to understand the concerns first.
Good fit when
- Reviews arrived, the deadline is short and it must be decided which experiments to run first.
- Three reviewers said different things and the concern behind each comment must be understood.
- New experiment results are ready and a response draft with manuscript revisions is needed from them.
- A finished response exists and should be checked for missed concerns and unsupported claims.
Not a fit when
- A ready response text for submission is wanted, and the skill only gives material for the author.
- The venue bans any generative assistance in replies to reviews.
- The reviews contain confidential chair notes that must not be passed to tools.
Example request
Analyze the reviews of my paper before drafting a response: infer the concern behind each comment, rate severity and build a P0 to P3 experiment plan for 5 days.Limitations
The skill prepares response material and gives no finished text: the author edits it and stays responsible for the correctness of methods, results and writing. The venue policy on generative assistance must be checked first. ACL and related conferences require disclosure when the use creates content. The skill uploads nothing by itself, but the review text goes to the agent model, so confidential material must not be passed to a chat. Stage 2 needs real experiment results and exact settings. The last commit in the repository is from August 2026.
How to disable. Delete the `Rebuttal-Skill` folder from the agent skills directory.
FAQ
How do I set up Rebuttal Skill?
The repository holds a skill definition. There is no application. Copy `SKILL.md` into the skills directory of an agent that supports skills, such as OpenCode, Claude Code or Gemini CLI.
Does the skill write a ready response to reviewers?
No. The skill produces material for the author: an analysis, a plan, a draft response structure and a revision list. The author rewrites the final text.
What inputs does the skill ask for?
Stage 1 needs the abstract, review texts, scores, venue and scale, deadline and available compute. Stage 2 adds experiment results, their exact settings and manuscript locations.
What if the chances look low?
Stage 0 marks the case `LOW EXPECTED RETURN` and builds a resubmission roadmap instead, with a rejection diagnosis and an experiment plan for the next version.
How does the skill treat venue rules?
Before drafting it checks the venue policy on generative assistance, flags required disclosure and leaves the final response with the author.
Discussion
No questions or reviews yet.
Related
SkillsA skill that turns a vague interest, a literature gap or a proposal draft into a testable research question with a first-step plan
A set of 16 skills for research work: ideation, experiments, paper writing, review, rebuttal, slides and research memory.
A skill that analyzes materials and subtitles, checks facts against sources and keeps a personal library in a local folder.
A collection of skills for decisions and projects: strategy, market research, thinking, data analysis, docs and slides.
A skill that builds one preparation document for a job interview: company research, job description, resume and likely questions.
An analytical skill that tests whether users will find value in a product idea: scenario, conditions, timing and delivery, with solidity ratings
README in short
A skill guide for working with reviews: a two-stage analysis, a viability check and a ban on inventing results. The response stays with the author.