Paper Write SCI

Collaboration Agreement with bensz-collect-bugs

For bugs caused by design flaws of this skill, first use
```
bensz-collect-bugs
```
to record them in
```
~/.bensz-skills/bugs/
```
in a standardized manner; do not directly modify the locally installed skill source code of the user; if there is a workaround, record the bug first, then continue to complete the task
Only when the user explicitly requests actions like "report bensz skills bugs" for public submission, use the local
```
gh
```
to upload new bugs to
```
huangwb8/bensz-bugs
```
; do not pull or clone the entire repository

This skill is used to write or optimize the main text of SCI journal papers based on LaTeX paper projects, Figure/Table annotations, and user supplementary requirements.

During execution, prioritize assigning deterministic steps to scripts and leaving heuristic judgments to AI:

Initialize workspace, normalize mode, select style, name plan files: use
```
config.yaml:scripts.prepare_workspace
```
Read long rule blocks from
```
references/
```
as needed; do not load all reference documents into the context at once

Objectives

Produce papers that are more like the author's own work rather than generic AI templates
Strictly protect consistency of numbers, logic, and terminology during writing and revision
Directly advance modifications by default; when the user requires human-machine collaboration, only output plans without directly modifying the paper
Converge all intermediate files into
```
<paper_dir>/.paper-write-sci/run_{timestamp}/
```
except for explicitly agreed external deliverables

Inputs

Input Item	Required	Description
Paper source code directory	Yes	Root directory of the LaTeX paper project
Figure/Table annotations	Yes	Explain what argument each chart supports and what key numbers it contains
User requirements	Optional	e.g., "Only modify Results", "Conservative polishing", "Strengthen Discussion"
Reference papers/author reference materials	Optional	Used to extract additional style signals; only learn the style, do not copy sentences
Running mode	Optional	Default values and aliases are based on `config.yaml:mode`
Style	Optional	Default values and available list are based on `config.yaml:style`

Outputs

autonomous

Directly modify the target main text files
Write analysis, review, and rendering logs into the current running directory
```
<paper_dir>/.paper-write-sci/run_{timestamp}/
```
If an available build chain is detected, attempt to re-render PDF and Word

collaborative

Only output plan files, with the filename pattern based on
```
config.yaml:runtime_outputs.collaborative_plan_pattern
```
, which defaults to including the current
```
run_id
```
Summarize paper defects, evidence, recommended repair solutions, affected files, and risks in the plan
Do not directly modify the paper content
Intermediate files other than the plan are still written to the current running directory
```
<paper_dir>/.paper-write-sci/run_{timestamp}/
```

Mode Rules

autonomous

(Default)

Fix problems directly when discovered
Adhere to the principle of minimal necessary changes
Must pass numerical review before writing any new numbers
Must pass final review of chapter responsibilities, full-text consistency, and logical tree before completion

collaborative

Read the entire paper first, then summarize defects and modification paths
The plan is only used as human review material, no changes are made to the main text
Plan filename, topic slug,
```
run_id
```
, and output directory are all based on
```
config.yaml:runtime_outputs
```
The plan content must include at least: problems, evidence, recommended actions, affected chapters, associated charts, style anchors, chapter division risks, risk explanations

Intermediate File Constraints

Except for the following "explicitly agreed external deliverables", all other intermediate files must be placed in

<paper_dir>/.paper-write-sci/run_{timestamp}/

```
plans/{collaborative_plan_pattern}
```
Final paper build products, such as existing PDF and Word outputs of the project

Prohibit using the old directories

.write-paper-sci/

and the earlier

.write-paper/

Style System

Style directory:
```
references/styles/
```
Style template:
```
references/styles/style-template.md
```
Default style is based on
```
config.yaml:style.default
```

Current Styles:

Style	Field	Function
`bensz-01`	Biomedical	Author-style centered, emphasizing problem orientation, method naming, strong numerical comparison, detailed figure legends friendly to non-field readers, and honest limitations
`general-01`	General SCI	Integrate official SCI writing suggestions, provide a robust default style when no field is specified

Style Usage Principles:

Prioritize using the style explicitly specified by the user
If the user provides reference papers or author materials, additional signals can be refined based on the selected style and saved to
```
analysis/reference-style.md
```
in the current running directory
If the user does not specify, use
```
config.yaml:style.default
```
Style serves to "make the paper more human-like", not to pile up fancy sentences
Style can only change the way of expression, not facts, numbers, or logic
When the task is "full polishing", "rewrite Discussion" or any request obviously involving de-resulting of
```
Discussion
```
, first use
```
general-01
```
to hold the bottom line of chapter division, then overlay the author sense of
```
bensz-01
```
; do not let the author style override chapter responsibilities

On-Demand Reference Files

Prioritize reading only the reference files required for the current task:

Style and chapter writing methods:

references/writing-style-guide.md

references/styles/bensz-01.md

references/styles/general-01.md

Execution guards and passing standards:
```
references/execution-guards.md
```

Collaborative plan template:

references/collaborative-plan-template.md

Numerical review template:

references/templates/number-check-template.md

Chapter responsibility review template:

references/templates/section-role-check-template.md

Discussion audit

template:

references/templates/discussion-role-check-template.md

Logical tree template:

references/templates/logic-tree-template.md

Logical review template:

references/templates/logic-check-template.md

Reading Recommendations:

Read
```
references/execution-guards.md
```
when involving chapter misplacement,
```
Discussion
```
rewriting, or full-text final review
Prioritize reading
```
writing-style-guide.md
```
and target style files when only doing stylized polishing or figure legend polishing
Read
```
references/styles/style-template.md
```
only when adding new styles

Workflow

Phase 0: Initialization

Confirm that the paper directory and Figure/Table annotations exist and are readable
Locate the main entry tex, prioritize checking
```
main.tex
```
Scan
```
\input{}
```
and
```
\include{}
```
to establish chapter path mapping
Identify the build method and protected format files

Run

python3 scripts/prepare_workspace.py --paper-dir <paper_dir> [--mode ...] [--style ...] [--topic ...] [--reference-material ...]

Confirm the current
```
run_{timestamp}
```
directory, collaborative plan path, and hidden workspace root directory based on the script output
If there are reference papers or author reference materials, refine supplementary style signals
If the task involves high-risk issues such as chapter responsibilities, numbers, abbreviations, or logic, read
```
references/execution-guards.md
```

Phase 1: Understand the Current Paper

Read the full main text and key front/back matter
Extract the argument supported by each chart by comparing with Figure/Table annotations
Identify existing high-quality paragraphs, weak paragraphs, factual risks, and style inconsistencies
Mark chapter responsibility risks, especially paragraphs where
```
Introduction
```
reveals results,
```
Results
```
over-interprets, and
```
Discussion
```
repeats results
Initially build a logical tree and write it to
```
analysis/logic-tree.md
```
in the current running directory
If the task involves full-text abbreviation management, establish
```
analysis/abbreviation-inventory.md
```
according to
```
references/execution-guards.md
```

Phase 2: Determine Execution Path

collaborative

Generate the plan file according to

config.yaml:runtime_outputs.collaborative_plan_pattern

The plan should focus on "defects and repair suggestions" rather than directly providing the revised full text
The plan must explicitly write out
```
section-role risk
```
Stop here, do not modify the main text

autonomous

Develop an internal revision order
Prioritize correcting factual errors, numerical risks, and logical breaks
Then perform structural optimization, language polishing, and style unification

Phase 3: Section-by-Section Writing or Revision

Default order:

text

Introduction -> Methods -> Results -> Discussion -> Conclusion ->
Additional Information -> Figure Legends -> Supplementary Materials -> Abstract

When processing each section:

First read the existing text and corresponding chart evidence
First judge the chapter responsibility of the section, then use the selected style to judge "how to write more like the author"
If long-rule judgment is involved, refer back to
```
references/execution-guards.md
```
If new numbers are involved, pass numerical review first
If the
```
Discussion audit
```
trigger condition is hit, complete the special review first, then start writing
If adding, deleting, or rewriting abbreviations is involved, check back
```
analysis/abbreviation-inventory.md
```
first
In
```
autonomous
```
mode, make minimal necessary modifications to the target tex and comply with
```
config.yaml:tex_readability
```
Update the logical tree and necessary full-text consistency records to ensure that local changes do not break the global consistency

When processing Figure Legends or Supplementary Materials:

Explain the minimal visual syntax required to read the figure for each panel
For terms or abbreviations that appear for the first time and are not intuitive enough, supplement the full name and a short definition
Only retain necessary method details that help read the figure, avoid repeating the entire Methods section into the legend

Phase 4: Final Review

Initiate
```
section-role-check
```
Complete full-text consistency review, including consistency of numbers, abbreviations, Figure/Table citations, and figure annotations
Rebuild the logical tree for the latest text and initiate logical review
If there are blocking issues, review again after fixing

Phase 5: Rendering and Delivery

Detect the build chain in the following priority:

Project-built one-click build or export script
```
scripts/manuscript_tool.py
```
```
Makefile
```
```
latexmk
```
Direct
```
xelatex
```

If a Word export chain exists, attempt to execute it as well. Put the build logs into

render/

of the current running directory.

Constraints

Do not modify

artifacts/

*.sty

*.cls

*.bst

*.bbx

*.cbx

latexmkrc

Do not modify the structure of
```
main.tex
```
unless explicitly requested by the user
Do not fabricate numbers
Do not downgrade abbreviation checks to "self-consistency in the current file"
Do not allow conflicting abbreviations, conflicting full names, or intermittent definitions of the same concept in different tex files
Do not arbitrarily create new abbreviations for local brevity; if the new abbreviation will not be stably reused in the full text, prefer not to use it
Do not rearrange irrelevant paragraphs or entire sections of
```
.tex
```
for "typesetting neatness"
Strictly limit the scope of changes when the user requests "only modify a certain section"
Strictly prohibit direct modification of the paper when the user requests collaborative mode

Reference Files

Workspace preparation script:
```
scripts/prepare_workspace.py
```
Execution guards:
```
references/execution-guards.md
```
Style template:
```
references/styles/style-template.md
```
```
bensz-01
```
:
```
references/styles/bensz-01.md
```

general-01

references/styles/general-01.md

Collaborative plan template:

references/collaborative-plan-template.md

Chapter writing guide:
```
references/writing-style-guide.md
```

Numerical review template:

references/templates/number-check-template.md

Chapter responsibility review template:

references/templates/section-role-check-template.md

Discussion audit

template:

references/templates/discussion-role-check-template.md

Logical tree template:

references/templates/logic-tree-template.md

Logical review template:

references/templates/logic-check-template.md

paper-write-sci

NPX Install

Tags

SKILL.md Content (Chinese)

Paper Write SCI

Collaboration Agreement with bensz-collect-bugs

Objectives

Inputs

Outputs

`autonomous`

`collaborative`

Mode Rules

`autonomous`
(Default)

`collaborative`

Intermediate File Constraints

Style System

On-Demand Reference Files

Workflow

Phase 0: Initialization

Phase 1: Understand the Current Paper

Phase 2: Determine Execution Path

`collaborative`

`autonomous`

Phase 3: Section-by-Section Writing or Revision

Phase 4: Final Review

Phase 5: Rendering and Delivery

Constraints

Reference Files