zrwangems/dsh-zrwang-plugin / ars/skills/academic-paper
academic-paper
Flagged学术论文写作:按 IMRaD 或领域结构成稿,含大纲、论证地图、双语摘要、图表与题注、引用格式转换、返修与逐条回复、AI 使用声明。Triggers: write a paper, academic paper, paper outline, revise paper, reviewer comments, rebuttal, citation check, abstract | 写论文, 论文大纲, 修改论文, 审稿意见, 返修, 审稿回复, 引文核对, 写摘要。
references/mode_selection_guide.md
View on GitHub> 输出语言(本 DSH 集成强制):所有面向用户的文字、稿件、摘要、评审与报告一律使用**简体中文**;英文术语与引文保持英文。文中的 `zh-TW` / `zh-tw-en` 只是上游语言代码与脚本参数,不代表要用繁体输出。用户明确要求其他语言时以用户为准。
>
> 多角色执行(评审团、十阶段流水线、角色扮演)必须使用 DSH 的 Agent Team 工具(`spawn_teammate`、`team_task_*`、`wait_agent`、`send_message`);当前会话看不到这些工具,说明用户没开 Agent Team,必须先提示用户开启,不得退化成单会话顺序扮演。
# Mode Selection Guide
This guide helps users and the `intake_agent` select the most appropriate operational mode.
---
## Mode Selection Flowchart
```
User Input →
│
├── Already have complete research?
│ ├── Yes → Want a full paper?
│ │ ├── Yes ─────────────────────────→ full mode
│ │ └── No → Just need an outline?
│ │ ├── Yes ─────────────────────→ outline-only mode
│ │ └── No → Just need an abstract?
│ │ ├── Yes ──────────────────→ abstract-only mode
│ │ └── No → Just need a literature review?
│ │ ├── Yes ─────────────→ lit-review mode
│ │ └── No ──────────────→ full mode
│ │
│ └── No → Want guided thinking?
│ ├── Yes ─────────────────────────→ plan mode ★ NEW
│ └── No ──────────────────────────→ full mode (Phase 0 will conduct an interview)
│
├── Have an existing paper to revise? ──────────────────────→ revision mode
├── Have comments explicitly identified as coming from a real committee or institutional review office?
│ └── Need tracking / response preparation ───────────────→ revision-coach committee-correspondence variant
├── Have reviewer comments to handle?
│ ├── Comments only, no response written yet ──────────→ revision-coach mode
│ └── Comments + an existing rebuttal/response draft ──→ rebuttal-audit mode
├── Need an AI-usage disclosure statement? ─────────────────→ disclosure mode
├── Just need format conversion? ────────────────────────→ format-convert mode
└── Just need a citation check? ────────────────────────→ citation-check mode
```
---
## Detailed Description of Each Mode
### full mode — Complete Paper Writing
**Applicable Scenarios**:
- User has a clear research question and (partial) materials
- Needs to produce a complete paper from start to finish
- Includes all phases: Interview → Literature → Structure → Argumentation → Writing → Citation → Review → Formatting
**Not Applicable When**:
- User has no idea about research direction (→ use `deep-research` first)
- Only need a specific section (→ use another specialized mode)
**Expected Output**: Complete paper draft + references + bilingual abstract + review report
**Expected Duration**: Long (all 8 Phases fully executed)
**Agents Used**: All 9 + socratic_mentor (if needed)
---
### outline-only mode — Outline Generation
**Applicable Scenarios**:
- Only need the paper structure and outline
- A proposal to submit to an advisor for review
- Need to quickly plan the paper structure
**Not Applicable When**:
- Need complete paper content (→ full mode)
- Need guided thinking (→ plan mode)
**Expected Output**: Detailed outline + evidence allocation + word count distribution
**Expected Duration**: Short (Phase 0-2)
**Agents Used**: intake → literature_strategist → structure_architect
---
### plan mode — Chapter-by-Chapter Guided Planning ★ NEW
**Applicable Scenarios**:
- User has ideas but they are not yet clear enough
- Wants guided thinking for each chapter's content
- First-time academic paper writer
- Wants to think through every section before writing
- Just received materials from deep-research and needs to transform them into a paper plan
**Not Applicable When**:
- Already knows exactly what to write (→ full mode is faster)
- Only needs an outline without deep thinking (→ outline-only mode)
- Time-pressured and needs rapid output (→ full mode)
**Expected Output**: Chapter Plan + INSIGHT Collection
**Expected Duration**: Medium (Step 0-3, approximately 20-30 rounds of conversation)
**Agents Used**: intake → socratic_mentor → structure_architect → argument_builder
**Subsequent Connections**:
- Chapter Plan → full mode (produce complete paper)
- Chapter Plan → academic-paper-reviewer (review the plan)
---
### revision mode — Paper Revision
**Applicable Scenarios**:
- Already have a completed paper draft
- Received reviewer comments requiring revision
- Feel certain sections need improvement
**Not Applicable When**:
- No existing paper draft (→ full mode)
- Only need to check citation format (→ citation-check mode)
**Expected Output**: Patch document + deterministically applied revised draft + apply report + revision log (#390) + Citation Audit Report with proposed citation fixes (#1068)
**Expected Duration**: Medium
**Agents Used**: peer_reviewer → draft_writer → citation_compliance (proposes citation fixes; does not edit the applied draft, #1068)
**Prerequisite**: User must provide existing paper content
---
### abstract-only mode — Abstract Writing
**Applicable Scenarios**:
- Paper is already complete, only need an abstract
- Need to submit a conference abstract
- Need a bilingual abstract
**Not Applicable When**:
- No paper content to summarize (→ full mode or plan mode)
**Expected Output**: Bilingual abstract in the run's declared output language pair (`output_language_pair`; default Chinese + EN) + keywords
**Expected Duration**: Short
**Agents Used**: intake → abstract_bilingual
---
### lit-review mode — Literature Review
**Applicable Scenarios**:
- Need a literature review on a specific topic
- Preparing the Literature Review chapter of a paper
- Need a systematic search strategy and literature matrix
**Not Applicable When**:
- Need a complete paper (→ full mode)
- Need an in-depth research investigation (→ deep-research)
**Expected Output**: Annotated bibliography + literature matrix + synthesis analysis
**Expected Duration**: Medium
**Agents Used**: intake → literature_strategist
---
### format-convert mode — Format Conversion
**Applicable Scenarios**:
- Already have paper content, need format conversion
- Markdown → LaTeX / DOCX / PDF
- Need to comply with a specific journal's formatting requirements
**Not Applicable When**:
- No existing content (→ full mode)
- Need content modifications (→ revision mode)
**Expected Output**: Document in target format
**Expected Duration**: Short
**Agents Used**: formatter used standalone
---
### citation-check mode — Citation Check
**Applicable Scenarios**:
- Already have a paper, only need to check citation format
- Final check before submission
- Switching citation format (e.g., APA → IEEE)
**Not Applicable When**:
- No existing citation list (→ full mode)
- Need to modify paper content (→ revision mode)
**Expected Output**: Citation error report + automatic correction suggestions
**Expected Duration**: Short
**Agents Used**: citation_compliance used standalone
---
### revision-coach mode — Reviewer-Comment Triage
**Applicable Scenarios**:
- Received reviewer comments / a decision letter and need them parsed into a structured plan
- Want a Revision Roadmap that classifies, maps, and prioritizes each comment
- Need a Response Letter Skeleton to start drafting replies
- Deciding whether to address or push back on a comment (disagreement posture)
- Non-journal scopes: conference rebuttal, grant-panel response, transfer-after-review
**Not Applicable When**:
- Already have a written rebuttal/response draft you want QA'd (→ rebuttal-audit mode)
- The paper itself needs rewriting based on the comments (→ revision mode)
**Expected Output**: Revision Roadmap + optional Tracking Template + Response Letter Skeleton
**Expected Duration**: Short-Medium
**Agents Used**: revision_coach_agent used standalone (no prior pipeline execution required)
**Committee-correspondence variant**: when the user explicitly identifies comments
from a real committee or institutional review office, this same entrypoint loads
`committee_correspondence_protocol.md` and emits the separate #668 byte-accounted
concern tracker plus placeholder response skeleton. It does not emit the reviewer
Roadmap, Schema 11, priority/severity, a determination, or a resolution claim. Formal
tone alone never activates the variant.
---
### disclosure mode — AI-Usage Disclosure Outcome
**Applicable Scenarios**:
- Paper is drafted and you need venue-specific disclosure text or a policy action checklist
- Submitting under a covered policy target (ICLR, NeurIPS, Nature, Science, ACL, EMNLP, or a covered medical-publishing target: ICMJE, NEJM, The Lancet, JAMA, BMJ, PLOS, Frontiers, publisher-wide 中华护理杂志社, journal-level 国际眼科杂志)
- Need placement/action guidance for every required manuscript or submission channel
**Not Applicable When**:
- No paper drafted yet — disclosure is a finishing step (→ full mode first)
- The venue is not in the policy database (confirm the venue's current policy manually)
**Expected Output**: Default venue path — `REQUIRED` / `ACTION_ONLY` / `NOT_REQUIRED` / `UNKNOWN` applicability plus typed halt status; policy-anchor path — anchor-specific render
**Expected Duration**: Short
**Agents Used**: formatter agent's standalone disclosure branch, which loads the disclosure protocol before the venue policy database or policy-anchor lookup; it does not run normal Phase 7 formatting
---
### rebuttal-audit mode — Rebuttal-Draft QA
**Applicable Scenarios**:
- You already wrote a rebuttal/response-to-reviewers draft and want it checked before sending
- Need a per-comment coverage table (which reviewer concerns the draft addresses, partially addresses, or misses)
- Want risk flags for tone (too combative), unsupported claims, or a response that misreads the reviewer's point
**Not Applicable When**:
- You have reviewer comments but have **not** written a response yet (→ revision-coach mode, which *generates* a skeleton)
- The manuscript itself needs revising (→ revision mode)
**Input gate**: requires BOTH (a) the reviewer comments / decision letter AND (b) an existing rebuttal draft. Comments only → route to revision-coach. If intent is ambiguous, clarify rather than guess.
**Expected Output**: Advisory QA report — per-comment coverage + gap list + risk flags. Generates no new response.
**Expected Duration**: Short
**Agents Used**: revision_coach_agent's comment-parsing capability, run standalone (advisory)
**Integrity boundary**: a standalone invocation runs **outside** the pipeline, so rebuttal-audit does **not** emit a Schema 11 ledger, does **not** write the Material Passport, and does **not** mark anything `ready_to_submit` or verified. The output is advisory QA only.
---
## Paths from deep-research
```
deep-research completed
│
├── deep-research (full mode) outputs:
│ RQ Brief + Methodology Blueprint + Annotated Bibliography + Synthesis Report
│ │
│ ├── Want to write the paper directly ──→ academic-paper (full mode)
│ │ intake_agent auto-detects materials, skips redundant questions
│ │
│ └── Want to plan before writing ──→ academic-paper (plan mode)
│ socratic_mentor leverages existing materials to accelerate guidance
│
└── deep-research (socratic mode) outputs:
INSIGHT Collection + Synthesis Report
│
├── INSIGHTs are sufficiently clear ──→ academic-paper (full mode)
│
└── Need more guidance ──→ academic-paper (plan mode)
socratic_mentor continues deepening from INSIGHTs
```
## Connecting to academic-paper-reviewer
```
academic-paper completed
│
├── full mode produces complete paper ──→ academic-paper-reviewer (full / guided)
│ Complete peer review + revision suggestions
│
├── plan mode produces Chapter Plan ──→ academic-paper-reviewer (guided)
│ Review the plan's feasibility and completeness
│
└── reviewer feedback ──→ academic-paper (revision mode)
Revise paper based on review comments
```
---
## Common Misselection Scenarios
| User Says | Easily Misselected | Correct Choice | Reason |
|---------|---------|---------|------|
| "Help me write an outline" / 「幫我寫大綱」 | outline-only | First confirm: Do they want a simple outline or deep planning? | May need plan mode |
| "I want to write a paper but don't know how to start" / 「想寫論文但不知道怎麼開始」 | full | plan mode | Needs guided thinking |
| "Help me revise my paper" / 「幫我修改論文」 | revision | First confirm: Are there reviewer comments? | May need full mode rewrite |
| "Help me search for literature" / 「幫我找文獻」 | lit-review | First confirm: Is it a literature review for a paper or a research investigation? | May need deep-research |
| "I have deep-research results, help me write a paper" / 「我有研究結果,幫我寫成論文」 | full (skip Phase 0 directly) | full (but intake needs to detect handoff) | Materials need to be properly imported |
| "I want to plan my paper step by step" / 「我想逐步規劃論文」 | outline-only | plan mode | Needs interactive guidance |
| "The paper format is wrong" / 「論文格式不對」 | revision | citation-check or format-convert | May only need format correction |
| 「帶我寫論文」/「引導我寫論文」 | full | plan mode | 使用者需要互動式引導,不是直接產出 |
| 「第一次寫論文」/「論文新手」 | full | plan mode | 新手需要蘇格拉底式逐章引導 |
---
## Quick Decision Table
| What Do You Have? | What Do You Want? | Choose This Mode |
|-----------|-----------|------------|
| Nothing | Complete paper | plan mode → full mode |
| Research question + literature | Complete paper | full mode |
| Research question + literature | Outline | outline-only mode |
| Vague idea | Paper plan | plan mode |
| deep-research results | Complete paper | full mode (auto-handoff) |
| deep-research results | Guided planning | plan mode |
| Completed paper | Revision | revision mode |
| Completed paper | Abstract | abstract-only mode |
| Completed paper | Format conversion | format-convert mode |
| Completed paper | Citation check | citation-check mode |
| Reviewer comments (no response yet) | Parse + roadmap + reply skeleton | revision-coach mode |
| Real-committee comments (explicitly identified; no response yet) | Preserve source + concern tracker + placeholder response skeleton | revision-coach committee-correspondence variant |
| Reviewer comments + a written rebuttal draft | QA the draft before sending | rebuttal-audit mode |
| Drafted paper + target venue | AI-usage disclosure bundle or policy action checklist | disclosure mode |
---
### Plan to Full Mode Conversion Protocol
When a user completes `plan` mode and wants to proceed to `full` mode for actual paper writing:
#### Conversion Checklist
| Plan Mode Output | Full Mode Input | Conversion Action |
|-----------------|-----------------|-------------------|
| Chapter Plan (structure outline) | `structure_architect` agent | Map chapters → formal sections with heading levels; validate against `paper_structure_patterns.md` |
| Socratic Responses (Q&A transcripts) | `argument_builder` agent | Extract claims + evidence + warrants from dialogue; discard conversational scaffolding |
| Literature Notes (if any) | `literature_strategist` agent | Independent execution — plan mode notes serve as seed keywords only; full systematic search required |
| Argument Sketches | `argument_builder` agent | Evaluate each sketch against 4-level scoring; only `adequate` or above proceed |
#### Quality Gate
Before conversion, ALL of the following must be true:
- [ ] Every chapter in the Chapter Plan has at least 1 argument sketch rated `adequate` or above
- [ ] The overall paper structure maps to a recognized pattern in `paper_structure_patterns.md`
- [ ] At least 5 potential references have been identified (seeds for `literature_strategist`)
- [ ] The research question is finalized (not still evolving from Socratic dialogue)
#### What Gets Discarded
- Conversational filler from Socratic dialogue (greetings, confirmations, repetitions)
- Tentative ideas explicitly marked as "maybe" or "not sure" by the user
- Plan mode's iterative drafts (only the final version of each chapter plan carries over)
---
## Trigger-to-Mode Mapping Examples
```
"Write a paper on SDGs in HEI" -> full
"Give me a paper outline for..." -> outline-only
"Revise this paper based on feedback" -> revision
"Write an abstract for this paper" -> abstract-only
"Do a literature review on..." -> lit-review
"Convert this paper to LaTeX" -> format-convert
"Convert citations to IEEE" -> format-convert
"Check the citations in this paper" -> citation-check
"guide my paper" -> plan
"help me plan my paper" -> plan
"I got reviewer comments" -> revision-coach
"parse these reviews" -> revision-coach
"help me with my revision" -> revision-coach
"should we push back on reviewer 2" -> revision-coach
"conference rebuttal" / "grant response" -> revision-coach
"track these committee comments" -> revision-coach committee-correspondence variant (only with explicit real-source identity)
"audit my rebuttal draft" -> rebuttal-audit (needs comments + an existing draft)
"did I miss any reviewer comment" -> rebuttal-audit
"AI disclosure for Nature" -> disclosure
"generate an AI usage statement" -> disclosure
```
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |