@researai/deepscientist 1.5.16 → 1.6.0

package/src/skills/idea/references/research-history-playbook.md

@@ -0,0 +1,114 @@
+# Research History Playbook
+
+Use this reference when the `idea` stage needs broad, history-aware literature search rather than a shallow scan of the last few papers.
+
+The goal is to move from a field map to a lineage view and then to the real unresolved breakpoints.
+
+## 1. Build a field map first
+
+Before deep reading, map the landscape:
+
+- main problem
+- major subproblems
+- representative paradigms
+- common datasets and metrics
+- recurring disputes or failure modes
+
+For a broad topic, start with:
+
+- `3` strong reviews, tutorials, or surveys
+- about `10` recent representative papers
+- about `5` seminal or foundational papers
+
+Do not start with deep reading of one narrow recent cluster and mistake that for the field.
+
+## 2. Find the three paper roles
+
+For any serious direction, identify at least:
+
+- seminal or foundational papers
+- paradigm-shift or turning-point papers
+- current SOTA or mainstream papers
+
+Only reading recent papers makes it too easy to confuse "currently fashionable" with "structurally important".
+
+## 3. Use citation chaining aggressively
+
+History reconstruction should use both:
+
+- backward citation chaining
+  - what did the key paper cite?
+- forward citation chaining
+  - who later cited the key paper?
+
+Use citation chaining to answer:
+
+- how was the problem originally defined?
+- which assumptions survived over time?
+- which assumptions changed?
+- which assumptions were never seriously tested?
+
+If DeepXiv is available under the runtime contract, prefer it for broad paper-centric discovery, citation expansion, and shortlist triage.
+If DeepXiv is unavailable, use search engines, citation expansion, arXiv discovery, and open-web search directly.
+
+## 4. Keep a lineage table, not just paper titles
+
+For the strongest `20-40` papers in the active neighborhood, track at least:
+
+- paper
+- research problem
+- core assumption or mechanism
+- method and data
+- main conclusion
+- explicit limitation
+- implicit limitation you detect
+
+This table is often where the real idea appears:
+
+- overworked problems
+- default assumptions with weak evidence
+- conclusions tied to one narrow data regime
+- real gaps versus fake gaps
+
+## 5. Search broadly but with structure
+
+Use a broad search ladder:
+
+1. direct neighborhood
+   - same task, dataset, and metric
+2. failure-mode neighborhood
+   - same bottleneck, breakdown pattern, or contradiction
+3. mechanism neighborhood
+   - same core idea in the same or adjacent tasks
+4. cross-domain neighborhood
+   - structurally similar solutions from nearby fields
+
+Use DeepXiv when available for broad paper-centric discovery.
+If it is not available, use search engines directly and keep the same breadth target.
+
+Broad search does not mean random search.
+It means wide enough to expose the main paradigms, history, and unresolved contradictions before selecting a route.
+
+## 6. Distinguish review intents
+
+Use the right review style for the current need:
+
+- narrative review
+  - good for understanding and viewpoint formation
+- scoping review
+  - good for broad boundary mapping and gap discovery
+- systematic review
+  - good for tighter, reproducible evidence synthesis on a focused question
+
+The `idea` stage usually begins closer to field mapping and scoping, then narrows toward targeted comparison once the frontier becomes smaller.
+
+## 7. Stop when the breakpoints are clear
+
+The history pass is good enough when you can answer:
+
+- what the field now treats as the mainline
+- what its strongest assumptions are
+- where the main contradictions, blind spots, or failure regions remain
+- why the current candidate still matters despite the strongest nearby work
+
+If these are still unclear, do not rush into novelty claims.

package/src/skills/idea/references/selection-gate.md

@@ -2,7 +2,35 @@
 
 Use this reference when choosing the final idea and preparing the handoff to `experiment`.
 
-## 1. Lightweight quality gate
+## 1. Value and feasibility screen
+
+Before promotion, score each serious candidate on a compact `1-5` scale:
+
+- importance
+- novelty
+- feasibility
+- verifiability
+- paper or report potential
+- failure value
+- breakthrough or boundary-changing potential
+
+Also check the FINER-style screen explicitly:
+
+- `F`
+  - feasible with current data, compute, codebase, and schedule
+- `I`
+  - interesting enough that the field would care
+- `N`
+  - novel in a meaningful sense relative to the strongest nearby work
+- `E`
+  - ethically acceptable and not obviously high-risk
+- `R`
+  - relevant to an important bottleneck rather than a decorative tweak
+
+If the route scores poorly on both value and feasibility, do not promote it merely because it feels exciting.
+If the route is only a small local tweak, require an explicit justification for why a more differentiated route did not survive.
+
+## 2. Lightweight quality gate
 
 Score the final serious candidate on a `0/1/2` scale:
 
@@ -31,20 +59,41 @@ If total `< 7/10`, do not promote the idea yet.
 Also treat these as hard gates before promotion:
 
 - the literature survey must already durably cover at least `5` and usually `5-10` related and usable papers
+- the survey must show both the strongest direct-field overlap and any adjacent-domain mechanisms that could plausibly subsume or outgrow the candidate
+- a pre-idea draft or equivalent durable challenge memo must already exist for the serious surviving candidates, especially the likely winner
+- that draft must explicitly surface hidden assumptions, local-optimum lock-in risk, strongest rejection case, and strongest outside-family alternative
 - the closest-prior-work comparison must explain why the idea is still needed
+- the selected route must be classified honestly as breakthrough-seeking, clearly differentiated, or incremental-with-justification; default rejection applies to decorative tweaks
 - the final selected-idea draft must be ready to carry standard-format citations for the papers actually used
 
-## 2. Honest novelty / value labels
+## 3. Honest novelty / value labels
 
 Use these labels explicitly:
 
-- `novel`: closest prior work does not already make the same mechanism-plus-claim combination
+- `novel`: closest prior work does not already make the same mechanism-plus-claim combination, and the route is materially more than a cosmetic delta from the baseline
 - `incremental but valuable`: overlap exists, but the new setting, evidence package, or failure-mode resolution is still meaningful
 - `not sufficiently differentiated`: closest prior work already dominates the idea
 
 Only the first two are eligible for promotion.
+Even then, prefer `novel` when it remains feasible.
+
+## 4. Mechanism and falsification gate
+
+Before a candidate can be promoted, it should make explicit:
+
+- core hypothesis
+- mechanism sketch
+- strongest falsification experiment
+
+The mechanism sketch can be brief, but it must answer:
+
+- why should this route work at all?
+- what part of the current limitation does it change?
+- for whom, where, or under what condition should it work or fail?
+
+If the mechanism sketch or strongest falsification experiment is still vague, the route is not yet ready.
 
-## 3. Handoff fields for experiment
+## 5. Handoff fields for experiment
 
 The selected idea record should include:
 
@@ -69,7 +118,7 @@ Inside the implementation handoff, also include:
 - `abandon_condition`
 - `strongest_alternative_hypothesis`
 
-## 4. Recommended presentation shape
+## 6. Recommended presentation shape
 
 Use a compact Pyramid structure:
 
@@ -78,14 +127,17 @@ Use a compact Pyramid structure:
 - then the minimal validation plan
 - then a short `References` or `Bibliography` section that cites the survey-stage papers actually used
 
-## 5. Promotion gate
+## 7. Promotion gate
 
 Do not promote a candidate if any of these remain unclear:
 
 - what exactly is being claimed
 - how the claim differs from closest prior work
+- what hidden assumption is carrying most of the route
+- whether the route only looks best because of incumbent inertia or implementation convenience
 - what minimal experiment can refute it
 - which code touchpoints are affected
 - what evidence package would later defend it in writing
 - which `5-10` surveyed papers actually support the motivation, mechanism, and claim boundary
 - whether the final idea draft includes proper citation markers plus a standard-format reference list
+- why this route deserves promotion over simpler but lower-value tweaks and over more ambitious but still feasible alternatives

package/src/skills/intake-audit/SKILL.md

@@ -7,6 +7,7 @@ skill_role: companion
 # Intake Audit
 
 Use this skill when the quest already has meaningful state and the first job is to normalize that state instead of restarting the canonical research loop from zero.
+The goal is to recover one trustworthy starting state from messy existing assets, not to re-audit everything forever.
 
 ## Interaction discipline
 
@@ -23,6 +24,12 @@ Use this skill when the quest already has meaningful state and the first job is
 - **For git inspection or maintenance inside the current quest repository or worktree, prefer `artifact.git(...)` before raw shell git commands.**
 - **Use shell execution only when durable quest files, artifacts, and memory are insufficient; do not bypass durable state just because shell feels faster.**
 
+## Three-layer todo contract
+
+- treat quest-root `plan.md` as the top-level research map whose next active node must become explicit after intake
+- if the audit is multi-step, use workspace `PLAN.md` as the current intake-node contract and `CHECKLIST.md` as the execution frontier
+- when the audit resolves the route, update quest-root `plan.md` instead of leaving the recommendation only in a report artifact
+
 ## Purpose
 
 `intake-audit` is an auxiliary entry skill, not a normal long-running anchor.
@@ -34,6 +41,9 @@ Its purpose is to answer four questions before deeper work begins:
 3. what can be reused directly?
 4. which skill should take over next?
 
+In practice, `intake-audit` should usually leave behind one authoritative current-board surface.
+That board packet exists so later `decision`, `idea`, `experiment`, and `write` passes do not have to reconstruct the active mainline from several partially stale state sources.
+
 This skill exists because many quests do **not** start from a clean slate.
 Common non-blank starts include:
 
@@ -72,6 +82,8 @@ They are common research entry states.
 - Do not mark an existing experiment as a durable main result unless it is genuinely the main run for an accepted idea line.
 - Do not silently import old drafts, plots, or notes as the active contract if they belong to a different idea line or branch line.
 - Do not lose provenance. If an artifact is reused, record where it came from and why it is trusted enough.
+- Do not confuse provenance with manuscript content. User requests, agent decisions, branch/worktree names, command logs, and restart notes are control evidence only unless rewritten into a neutral scientific protocol by `write`.
+- Do not import old paper matrices or outline rows as current-method support until legacy-method, comparator, negative-evidence, appendix-only, and latest-method roles are separated.
 - If the quest is really a review/revision task, route to `rebuttal` instead of pretending this is a normal fresh paper-writing pass.
 
 ## Typical intake states
@@ -95,6 +107,14 @@ Also classify every important asset by trust:
 - `stale_or_conflicting`
 - `missing_context`
 
+Also classify every paper-facing asset by manuscript visibility:
+
+- `main_text_candidate`
+- `appendix_or_reproducibility`
+- `comparator_or_negative_evidence`
+- `reference_only`
+- `internal_only`
+
 ## Primary truth sources
 
 Use, in roughly this order:
@@ -199,7 +219,24 @@ Then reconcile it with the durable artifact layer:
 
 If the evidence is insufficient for a durable backfill, record that insufficiency explicitly instead of inventing a cleaned-up history.
 
-### 5. Choose the next anchor
+### 5. Compile the current board packet
+
+Before choosing the next anchor, compress the intake result into one durable current-board packet.
+
+At minimum, make explicit:
+
+- `current_mainline`
+- `incumbent`
+- `latest_decisive_result`
+- `active_blocker`
+- `stale_routes_to_ignore`
+- `next_decision_scope`
+- `budget_class`
+
+The point is not to summarize everything.
+The point is to give the next skill one authoritative board surface instead of forcing it to merge branch state, memory, summaries, and artifacts again from scratch.
+
+### 6. Choose the next anchor
 
 After reconciliation, write one durable route decision with `artifact.record(payload={'kind': 'decision', ...})`.
 
@@ -212,7 +249,7 @@ Typical next anchors:
 - review package is active -> `rebuttal`
 - the quest is effectively complete or should pause -> `finalize`
 
-### 6. Report and hand off
+### 7. Report and hand off
 
 At the end of the intake pass, send one threaded `artifact.interact(kind='milestone', ...)` update that says:
 
@@ -224,6 +261,7 @@ At the end of the intake pass, send one threaded `artifact.interact(kind='milest
 ## Recommended durable outputs
 
 - `artifacts/intake/state_audit.md`
+- `artifacts/intake/current_board_packet.md`
 - `artifacts/intake/recommended_next_step.md`
 - one `decision` artifact for the post-audit route
 - one or more repair/backfill artifact calls when justified
@@ -275,5 +313,8 @@ When the audit concerns a specific existing line, include identifiers when known
 - the quest's current state is understandable
 - the trustworthy reusable assets are explicit
 - the untrusted gaps are explicit
+- the current board packet is explicit enough that a later skill can continue from one authoritative board surface
 - the next anchor is explicit
 - the system can continue without pretending the quest started from zero
+
+A good intake pass tells the system exactly what state is trusted, what is not, and what runs next.

package/src/skills/intake-audit/references/state-audit-template.md

@@ -33,6 +33,16 @@
 - missing metric contract:
 - stale draft risk:
 
+## Current Board Packet
+
+- current_mainline:
+- incumbent:
+- latest_decisive_result:
+- active_blocker:
+- stale_routes_to_ignore:
+- next_decision_scope:
+- budget_class:
+
 ## Route Recommendation
 
 - next anchor:

package/src/skills/nature-data/SKILL.md

@@ -0,0 +1,128 @@
+---
+name: nature-data
+description: >-
+  Prepare, audit, or revise Nature-ready Data Availability statements, data repository plans,
+  dataset citations, and FAIR metadata checklists for manuscripts. Use when the user asks about
+  Nature data availability, research data sharing, repository selection, accession numbers,
+  restricted or sensitive data, source data, supplementary datasets, DataCite-style dataset
+  references, FAIR metadata for academic publication, or Chinese-to-English data availability
+  wording for Chinese-speaking authors preparing Nature-family submissions.
+skill_role: companion
+---
+
+# Nature Data Availability Skill
+
+This companion skill is adapted from `Yuan1z0825/nature-skills/tree/main/nature-data`.
+See `UPSTREAM_LICENSE.txt` for the upstream MIT license.
+
+## DeepScientist integration
+
+- Follow the shared interaction contract injected by the system prompt.
+- Use this as a focused writing companion inside `write`, `review`, `rebuttal`, or `finalize` when the active issue is Data Availability, source data, repository selection, dataset citation, restricted data, or FAIR metadata.
+- Keep DeepScientist's evidence contract authoritative: draft availability text only from verified data inventory, repository records, artifact paths, or explicit unresolved fields.
+- Do not invent DOIs, accession numbers, repositories, ethics approvals, access committees, licences, embargoes, or data-use conditions.
+
+Use this skill to turn a manuscript's supporting data into a transparent, Nature-ready data
+availability package: statement text, repository plan, dataset citations, and missing-information
+flags.
+
+The governing policy layer is Springer Nature / Nature Portfolio data policy. The implementation
+layer is FAIR data practice and DataCite-style citation metadata.
+
+## Chinese-user operating mode
+
+When the user writes in Chinese, provides a Chinese manuscript note, or asks for "中文对应",
+"中英对照", "数据可用性声明", "数据获取声明", "原始数据", "数据存储库", or "受限数据":
+
+- Accept Chinese input naturally, but draft the final submission-ready statement in English unless
+  the user explicitly asks for Chinese only.
+- Preserve a short Chinese explanation of unresolved decisions when it helps the author act.
+- Translate intent, not wording. Chinese phrases such as "可向通讯作者索取" are usually too vague
+  for Nature-style English unless the restriction and access process are specified.
+- Convert Chinese repository/status descriptions into precise publication terms:
+  `数据可用性声明` -> `Data Availability`; `原始数据` -> `raw data`;
+  `处理后数据` -> `processed data`; `源数据` -> `source data`;
+  `补充材料` -> `Supplementary Information`; `受限数据` -> `restricted data`;
+  `合理请求` -> `reasonable request`, only with reason and review route.
+- Use `references/chinese-author-alignment.md` for Chinese terminology, common CN-to-EN failure
+  modes, and bilingual intake questions.
+
+## Default stance
+
+- Treat the Data Availability statement as a link between the paper's claims and the evidence
+  needed to inspect, reproduce, or reuse them.
+- Do not invent DOIs, accession numbers, repository names, licences, embargo dates, ethics
+  approvals, access committees, or data-use conditions.
+- Prefer public, discipline-specific repositories. Use generalist or institutional repositories
+  only when no suitable community repository exists.
+- Describe both newly generated data and reused third-party data.
+- If data cannot be openly shared, state why, who controls access, how requests are evaluated,
+  and what metadata or representative data can still be public.
+- Separate data, code, materials, and protocols unless the journal asks for a combined
+  availability section.
+- Keep this skill focused on availability and metadata. Do not rewrite methods, analyze
+  statistics, or polish the manuscript unless the user asks for those tasks separately.
+- Flag "available upon request" as weak unless there is a specific legal, ethical, commercial, or
+  third-party restriction.
+
+## Workflow
+
+1. Identify the target journal and article type. If journal-specific instructions conflict with
+   this skill, follow the journal.
+2. Inventory every dataset needed to support the main and supplementary results:
+   generated raw data, processed data, figure source data, secondary data, software outputs,
+   models, tables, images, and files underlying statistical analysis.
+3. Classify each dataset into one access route:
+   `public repository`, `controlled access repository`, `within paper or supplement`,
+   `reused public source`, `third-party restricted`, `available on justified request`,
+   or `not applicable`.
+4. Choose repository and identifier strategy before drafting text. Prefer DOI, accession number,
+   Handle, ARK, or stable repository record over personal websites and temporary cloud links.
+5. Draft the Data Availability statement using explicit dataset-to-location mapping.
+6. Add formal dataset citations for public data that support conclusions.
+7. Run the FAIR and metadata audit before finalizing.
+8. Return ready-to-paste statement text plus any unresolved fields the author must confirm.
+
+## Output format
+
+Unless the user asks for another format, return:
+
+```text
+Data Availability
+[ready-to-paste statement]
+
+Repository and citation actions
+- [specific actions or "None"]
+
+Missing information / risk flags
+- [specific flags or "None"]
+
+中文核对
+- [用中文列出作者需要确认的字段或 "无"]
+```
+
+When auditing an existing statement, lead with blocking issues first, then provide a revised
+version.
+
+## Related files
+
+| File | Open when |
+|---|---|
+| [references/policy-principles.md](references/policy-principles.md) | You need the governing Nature/Springer Nature data-sharing rules or edge-case policy logic |
+| [references/chinese-author-alignment.md](references/chinese-author-alignment.md) | The user writes in Chinese, needs bilingual wording, or provides Chinese availability notes |
+| [references/statement-patterns.md](references/statement-patterns.md) | You need ready-to-adapt Data Availability statement patterns |
+| [references/repository-and-identifiers.md](references/repository-and-identifiers.md) | You need repository choice, accession, DOI, embargo, versioning, or dataset citation guidance |
+| [references/fair-metadata-checklist.md](references/fair-metadata-checklist.md) | You need FAIR checks, README metadata, file organization, licences, provenance, or DataCite fields |
+| [references/source-basis.md](references/source-basis.md) | You need to justify rules with official sources or check which source supports which rule |
+
+## Source hierarchy
+
+Use sources in this order:
+
+1. Target journal instructions and submission system requirements.
+2. Nature Portfolio / Springer Nature data, code, materials, and reporting policies.
+3. Repository-specific requirements and domain community standards.
+4. FAIR principles and DataCite metadata practice.
+
+If a policy detail may have changed, verify the current journal page before giving final
+submission advice.

package/src/skills/nature-data/UPSTREAM_LICENSE.txt

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Yuan Yizhe
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/src/skills/nature-data/agents/openai.yaml

@@ -0,0 +1,4 @@
+interface:
+  display_name: "Nature Data"
+  short_description: "Draft bilingual-aware Nature data statements"
+  default_prompt: "Use $nature-data to turn my Chinese or English data notes into a Nature-style Data Availability statement, repository plan, and FAIR metadata checklist."

package/src/skills/nature-data/references/chinese-author-alignment.md

@@ -0,0 +1,84 @@
+# Chinese Author Alignment
+
+Use this file when the user writes in Chinese, provides a Chinese Data Availability draft, or asks
+for bilingual wording. The goal is not to translate Chinese literally. The goal is to convert the
+author's Chinese description into a Nature-ready English availability route.
+
+## Core terminology
+
+| 中文 | Preferred English | Notes |
+|---|---|---|
+| 数据可用性声明 / 数据获取声明 | Data Availability | Use the journal heading `Data Availability`. |
+| 本研究产生的数据 | data generated in this study | Include repository and identifier when public. |
+| 原始数据 | raw data | Do not call processed tables raw data. |
+| 处理后数据 | processed data | State whether processing scripts are available. |
+| 源数据 | source data | Usually data underlying figures or tables. |
+| 补充材料 / 附录 | Supplementary Information | Use exact file/table names when possible. |
+| 公共数据库 | public database / public repository | Name the database and identifier. |
+| 数据存储库 | data repository | Prefer repository over platform unless it is a true archive. |
+| 登录号 / 编号 | accession number | Use for repositories that assign accession IDs. |
+| DOI / 永久链接 | DOI / persistent URL | Prefer DOI when available. |
+| 受限数据 | restricted data | Explain legal, ethical, consent, commercial, or third-party reason. |
+| 脱敏数据 | de-identified data | Do not say anonymous unless re-identification risk is addressed. |
+| 合理请求 | reasonable request | Not enough alone; add route, eligibility, and conditions. |
+| 通讯作者 | corresponding author | Avoid making an email the only durable access route if an institutional route exists. |
+| 数据使用协议 | data-use agreement | State when required for access. |
+| 伦理审批 | ethics approval | Name approval body or requirement when relevant. |
+| 代码可用性 | Code Availability | Keep separate if the journal separates data and code. |
+
+## Chinese-to-English conversion rules
+
+- Convert "本文所有数据均包含在正文和补充材料中" to a specific claim:
+  name Source Data files, Supplementary Tables, or repository records. If raw data are absent, say
+  so as a risk flag rather than pretending they are included.
+- Convert "可向通讯作者合理索取" only after adding:
+  why public sharing is impossible, who reviews requests, eligible requesters, required approvals
+  or data-use agreement, and expected access route.
+- Convert "数据因隐私原因不可公开" into a controlled-access pattern:
+  state privacy/consent/legal basis, public metadata if available, access committee or institution,
+  and conditions.
+- Convert "商业数据/企业数据不可公开" into a third-party or commercial restriction pattern:
+  name the provider or owner, request route, and whether derived or aggregate data can be shared.
+- Convert "数据将在接收后上传" into an action item:
+  deposit before submission or create a private reviewer link if the repository supports it.
+- Convert "使用公开数据集" into a citation requirement:
+  include source, version/release/date accessed when relevant, and dataset citation.
+
+## Bilingual intake questions
+
+Ask only what is needed for the statement.
+
+```text
+请确认这些字段：
+1. 哪些数据支撑主文图、补充图和统计分析？
+2. 每类数据是否已有仓库、DOI、登录号或审稿人私密链接？
+3. 是否包含人类参与者、隐私、商业、第三方授权或国家/机构限制？
+4. 如果数据不能公开，谁负责审核申请？需要伦理审批或数据使用协议吗？
+5. 是否有代码、脚本或 README 能解释 raw data 到 figure source data 的处理过程？
+```
+
+## Common Chinese draft fixes
+
+| 中文原意 | Avoid literal English | Nature-ready direction |
+|---|---|---|
+| 数据可向通讯作者索取。 | Data are available from the corresponding author upon request. | State the restriction reason and institutional access process. |
+| 所有数据见补充材料。 | All data are in the supplementary materials. | Name exact Supplementary Tables/Source Data and flag missing raw data if any. |
+| 数据暂未上传。 | Data will be uploaded later. | Deposit now or list repository action as blocking. |
+| 使用了公开数据库。 | Public databases were used. | Name database, accession/version/date accessed, and cite dataset. |
+| 因隐私不能公开。 | Data cannot be public for privacy reasons. | Add de-identification status, access committee, eligibility, and agreement terms. |
+
+## Recommended bilingual output
+
+When useful, provide English first and Chinese second:
+
+```text
+Data Availability
+[English statement for submission]
+
+中文核对
+- 这句话对应中文含义：[brief Chinese explanation]
+- 需要作者确认：[missing accession / repository / ethics condition]
+```
+
+Do not put Chinese explanatory notes inside the final English statement unless the target journal
+allows bilingual manuscript text.