# Omnirebuttal

> >-

- **Type:** Skill
- **Install:** `agentstack add skill-iayce-omnirebuttal-omnirebuttal`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [Iayce](https://agentstack.voostack.com/s/iayce)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [Iayce](https://github.com/Iayce)
- **Source:** https://github.com/Iayce/omnirebuttal

## Install

```sh
agentstack add skill-iayce-omnirebuttal-omnirebuttal
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Omnirebuttal

Turn reviewer feedback into a venue-correct, evidence-backed author response. **Router + shared
kernel + specialized artifact branches** — not a lowest-common-denominator merge.

## When to use

- Conference rebuttal (CV one-page PDF, OpenReview, ICML per-review text)
- Journal major/minor revision response letters (Nature-style point-by-point)
- ARR/TMLR rolling review responses
- Rebuttal strategy, Issue Board, or REVISION_PLAN without full draft (`triage-only`)

## Nine-step workflow

0. **Venue Research (mandatory)** — Read [references/intake-venue-research.md](references/intake-venue-research.md): search **official** rules + **community** tips (官网, OpenReview, 小红书, 知乎, Reddit, …) via **web-access**; write `VENUE_BRIEF.md` before drafting.
1. **Intake & Route** — Read [references/intake-routing.md](references/intake-routing.md): venue, artifact family, limits, task mode (informed by brief).
2. **Triage** — Read [references/triage-issue-board.md](references/triage-issue-board.md): Issue Board + merge map.
3. **Strategy** — Read [references/strategy-decision.md](references/strategy-decision.md): champion, color-code, postures; flag AC escalation / Official Comment need.
4. **Experiment plan (conditional)** — Read [references/experiment-rebuttal-router.md](references/experiment-rebuttal-router.md): classify each experiment ask (Tier A–D, feasibility); write `EXPERIMENT_PLAN.md`; author-confirm before any `RUN`.
5. **Draft** — Load **one** primary artifact from [references/artifacts/](references/artifacts/); add [official-comment.md](references/artifacts/official-comment.md) when strategy warrants.
6. **Compress & Fit** — Read [references/compression/space-engineering-router.md](references/compression/space-engineering-router.md); apply PDF or text compression; run **page-fill pass** for `ONE_PAGE_PDF`; run `scripts/count_limits.sh` when applicable.
7. **Citation verify (when flagged)** — Read [references/citation-verification-lightread.md](references/citation-verification-lightread.md): use **LightRead CLI (`lr`)** to validate corrected BibTeX before claiming fixes in rebuttal.
8. **Coverage audit** — Read [references/reviewer-coverage-audit.md](references/reviewer-coverage-audit.md): reviewer × concern × response map; score-lift gaps; optional 中文 author sign-off table.
9. **Safety + QA** — [references/safety-gates.md](references/safety-gates.md) + [references/qa-checklist.md](references/qa-checklist.md).

## Artifact family router

| Family | Venues (examples) | Artifact file | Output |
|--------|-------------------|---------------|--------|
| `ONE_PAGE_PDF` | CVPR, ICCV, ECCV | [one-page-pdf-cv.md](references/artifacts/one-page-pdf-cv.md) | 1-page `rebuttal.tex` / PDF |
| `PER_REVIEW_TEXT` | ICML, KDD | [per-review-text.md](references/artifacts/per-review-text.md) | Text per reviewer + char limit |
| `THREADED_DISCUSSION` | NeurIPS, ICLR, UAI, **ACM MM** | [threaded-openreview.md](references/artifacts/threaded-openreview.md) | OpenReview thread(s) |
| `JOURNAL_POINT_BY_POINT` | Nature, R&R journals | [journal-point-by-point.md](references/artifacts/journal-point-by-point.md) | Response letter + change map |
| `ROLLING_REVISION` | ARR, TMLR | [rolling-arr-tmlr.md](references/artifacts/rolling-arr-tmlr.md) | Response + revision plan |
| *(secondary)* Official Comment / AC note | When strategy flags escalation | [official-comment.md](references/artifacts/official-comment.md) | Restricted or public chair-facing note |

Venue details: [references/venue-matrix.md](references/venue-matrix.md).

## Progressive loading

| Need | Read |
|------|------|
| **Live venue rules + community tips** | intake-venue-research.md → `VENUE_BRIEF.md` |
| Venue / family / mode | intake-routing.md, venue-matrix.md |
| Official Comment / AC-only note | artifacts/official-comment.md |
| Issue Board | triage-issue-board.md |
| Champion, ordering, postures | strategy-decision.md |
| Experiment asks (run / clarify / defer) | experiment-rebuttal-router.md |
| Provenance / promises / coverage | safety-gates.md |
| One-page PDF layout, tiny tables, figs | compression/one-page-pdf-latex.md |
| OpenReview char budget, markdown tables | compression/text-budget-markdown.md |
| Table vs figure choice | compression/visual-density-patterns.md |
| Copy-ready phrases | templates/response-templates.md |
| ICCV 2025 win case | cases/fastjsma-iccv2025.md |
| Source provenance (web + SKMP) | external-sources.md |
| Citation fix (LightRead `lr`) | citation-verification-lightread.md |
| Reviewer coverage / 中文对照表 | reviewer-coverage-audit.md |
| Benchmark scale / QC diplomacy | phrasing/diplomatic-benchmark-rebuttal.md |

## Default stance (all families)

- **AC / editor first** — rebuttal may be all they read.
- **Don't promise—do** — show table, equation, or paragraph now; then note paper update.
- Every Issue Board item gets an anchor; merge duplicates via `@All` or `merge_key`.
- No invented numbers, citations, or manuscript edits (three safety gates).

## Iron rules — NEVER violate

> **REVIEWER MAP MUST NOT INCLUDE RATINGS OR CONFIDENCE SCORES.**
>
> The reviewer map line (e.g. `Reviewer map. R1=Lv29; R2=jPgC; ...`) must contain
> **only the reviewer ID/alias** — nothing else. Explicitly prohibited:
> - Rating labels: "Borderline", "Weak Reject", "Strong Accept", "Accept", "Reject", etc.
> - Confidence levels: "Expert", "Knowledgeable", "Familiar", "Basic", or any numerical score.
> - Any meta-information about the reviewer's score or expertise.
>
> Rationale: Including ratings in the rebuttal document is visible to reviewers and
> area chairs, and signals to reviewers how their score compares to others. This can
> undermine the review process and is considered unprofessional. The reviewer map line
> serves only as a colour-coding legend.
>
> Correct format: `Reviewer map. R1=Lv29; R2=jPgC; R3=ypyP; R4=fXF9.`  
> Wrong format:   `Reviewer map. R1=Lv29 (Expert, Borderline); R2=jPgC (Familiar, Weak Reject); ...`

## Standard output package

Unless the user asks otherwise:

```text
VENUE_BRIEF.md (official limits + community intel + sources)
EXPERIMENT_PLAN.md (when any reviewer asks for new numbers)
Response strategy summary
Comment tracker (Issue Board + merge map)
Draft rebuttal / response letter (venue-specific)
Official Comment / AC-only note (if warranted)
Manuscript follow-ups (REVISION_PLAN)
Risk flags / AUTHOR_INPUT_NEEDED
Reviewer coverage map (optional 中文)
中文核对 (optional)
```

## Tools

```bash
# Character limit check (OpenReview / ICML text)
scripts/count_limits.sh draft.txt --chars --limit 5000

# Citation verification (LightRead CLI — install: bun add -g lightread-cli)
lr auth status --verify --format json
lr web fetch "https://aclanthology.org/2024.findings-acl.300/" --format json
lr search "Paper Title" --format json
```

## Legacy skill

CV-only workflows previously used `conference-rebuttal-playbook` — now superseded by
`artifacts/one-page-pdf-cv.md` within this skill. Playbook remains as deprecated redirect.

## What this skill does NOT do

- Run experiments automatically
- Submit to OpenReview / CMT
- Replace venue-specific legal/policy advice — verify CFP before final submit

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [Iayce](https://github.com/Iayce)
- **Source:** [Iayce/omnirebuttal](https://github.com/Iayce/omnirebuttal)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-iayce-omnirebuttal-omnirebuttal
- Seller: https://agentstack.voostack.com/s/iayce
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.