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/committee_correspondence_protocol.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。
# Committee Correspondence Protocol (#668)
This is the `committee-correspondence` variant of `academic-paper`
`revision-coach`. It structures comments already returned by a real committee or
institutional review office. It never simulates a committee, predicts satisfaction,
or makes a human-subjects determination.
Canonical schema:
`_shared/contracts/human_subjects/committee_correspondence.schema.json`.
Artifact version: `committee-correspondence/1.0`.
Deterministic checker: `scripts/check_committee_correspondence.py`.
Design authority:
`docs/design/2026-08-08-668-committee-correspondence-spec.md`.
## Activation gate
Use this variant only when the user explicitly identifies the input as comments or
a letter from a real committee/institutional review office and asks for tracking,
response preparation, or resubmission organization. Do not infer official authority
from formal tone, letterhead-like prose, words such as “required,” or a file name.
If the source is ordinary manuscript peer review, use normal `revision-coach`.
Journal or conference reviewers, editors, area chairs, and program committees are
peer review, even when the user names the venue or the venue calls the role a
committee (#854). If the source identity is uncertain, ask the user to identify it
before choosing the variant. V1 accepts UTF-8 letter text. For a binary PDF/DOCX,
retain the user’s original file outside this bundle and obtain a UTF-8
export/transcription; label that textual source honestly rather than claiming byte
identity with the binary.
This variant is standalone and separate from paper-review traceability. It MUST NOT:
- emit Schema 11 `commitment_extracted` or a peer-review Revision Roadmap;
- assign Major/Minor/Editorial, P1/P2/P3, severity, or model priority;
- write to the Material Passport or claim a Stage 4.5 integrity pass;
- state that a concern is resolved, that a response will satisfy the committee, or
that any artifact is submission-ready;
- translate committee wording into another jurisdiction’s taxonomy.
## Required interaction
1. Preserve the supplied UTF-8 text exactly as `source_letter.txt`. Do not normalize
newlines, whitespace, punctuation, or spelling.
2. Segment every byte from zero through EOF into `comment` or `non_comment` rows.
Headers, sign-offs, decision metadata, and other non-comment text remain visible.
3. Show the byte-complete segmentation to the user. The user confirms or corrects
comment boundaries before the tracker is final.
4. Create one concern per confirmed comment segment in source order. A compound
comment remains one concern with multiple `action_type` values. Separately
numbered subcomments may be separate segments.
5. Copy `verbatim_text` from the exact source bytes. Compute all SHA-256 fields and
byte ranges; do not ask a model to reproduce hashes from memory.
6. Classify `authority_status` only from explicit committee language or author
confirmation. When neither exists, use `unclear` plus an unresolved basis.
7. Fill artifact, owner, dependency, and required-before fields only when the letter,
author, or selected profile supplies them. Empty/unresolved is valid and preferred
to inference.
8. Present the concern records to the user for confirmation. Working views may
reorder the complete concern set but never replace source order.
9. Emit the placeholder-only response skeleton. It contains one concern marker,
author-response placeholder, and evidence placeholder per concern.
10. Run the deterministic checker and report any failure without claiming partial
completion.
## Bundle layout
Use `committee_correspondence/<source_sha256_first_12>/`:
```text
<12-hex>/
├── source_letter.txt
├── concern_tracker.json
└── response_skeleton.md
```
All tracker paths are bundle-relative. Never point the checker at arbitrary local
files through a tracker path. Do not use symlinks.
## Classification vocabulary
`authority_status` is one of:
- `explicitly_required` — the source explicitly requires the action;
- `conditional` — the source states an if/then or contingent requirement;
- `question` — the source asks for information without settling the answer;
- `suggestion` — the source explicitly presents an optional suggestion;
- `unclear` — none of the above is textually supported or the user has not confirmed.
`action_type[]` is multi-valued:
- `design`
- `explanation`
- `revise_artifact`
- `add_artifact`
- `administrative`
- `legal_or_policy_check`
Do not collapse a compound comment into a single label. Do not add a priority or
severity field. Source order is not priority.
## Optional authority-profile enrichment
No selected profile is a supported terminal input state:
```json
{
"state": "not_selected",
"selected_profile_ids": [],
"artifact_resolution_state": "artifact_agnostic"
}
```
When a user later selects a #666 profile, profile-derived artifact names and
requirement ids carry explicit provenance. The profile cannot rewrite source text,
authority status, owner, deadline, dependency, or order. A failed lookup uses
`artifact_resolution_state=unresolved`; it does not silently fall back to another
jurisdiction.
## Response skeleton
The skeleton begins with exactly:
```markdown
Status: drafting aid — no concern is asserted resolved.
```
Each concern block uses:
```markdown
<!-- concern:CC-001 -->
### Concern CC-001
**Draft response:** [AUTHOR RESPONSE REQUIRED — drafting aid; do not claim resolved]
**Evidence:** [EVIDENCE OR ARTIFACT POINTER REQUIRED]
```
It ends with exactly:
```markdown
> **Human-subjects boundary:** This output does not authorize recruitment, consent, access to identifiable data, intervention, or data collection.
```
The skeleton is not a completed response and contains no claim that work occurred.
During later drafting, keep the original skeleton as the checked baseline; save a
drafted response as a separate artifact rather than overwriting its hash-bound
placeholder state.
## Handoff
After user confirmation and a passing checker result, hand off both
`concern_tracker.json` and `response_skeleton.md` to response drafting or
resubmission work. Downstream work joins on `concern_id` and preserves the source
locator. Evidence remains absent until the author supplies a real artifact pointer.
Run:
```bash
python scripts/check_committee_correspondence.py \
committee_correspondence/<12-hex>/concern_tracker.json
```
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 |