papersmart-init

name

papersmart-init

description

Initialize a PaperSmart scientific-manuscript workspace or a new PaperSmart project folder in English or Chinese, including the projects/shared directory model, required draft/reference/output/config/log templates, writing_samples, language profile, and companion skills. Use when the user asks to set up PaperSmart, create the base file system, choose English or Chinese workspace mode, scaffold a manuscript project, install dependencies, or prepare a reusable PaperSmart workspace for other users.

PaperSmart-init

Overview

Set up the PaperSmart working environment before any manuscript writing. Preserve the separation between paper-specific project files and reusable cross-project resources.

Use scripts/init_papersmart.py for first-time filesystem setup. For routine new papers after the workspace exists, use PaperSmart-new-project.

Language Mode

Before running initialization, ask the user which workspace language to use:

• English folders and templates: default.
• 中文文件夹和模板：用户明确选择中文时使用。

Store the choice in shared/memory/papersmart_profile.md for English mode or 共享/记忆/papersmart_profile.md for Chinese mode. Later PaperSmart skills must read this profile first and answer in the configured output language.

Inputs

• Workspace root, defaulting to the current directory.
• Optional project slug, for example paper_01_digital_doppelganger.
• Optional working title, target journal, manuscript language, and short research objective.
• Optional instruction to install companion skills immediately.

Directory Contract

Create this workspace shape:

projects/
  README.md
shared/
  README.md
  docs/
  prompts/
  tools/
  skills/
  memory/

When a project slug is provided, create:

projects/<project_slug>/
  README.md
  config/project_config.md
  logs/writing_log.md
  logs/decision_log.md
  01_draft/README.md
  01_draft/data_inventory.md
  01_draft/data/
  01_draft/figures/
  01_draft/tables/
  01_draft/notes/
  02_reference/README.md
  02_reference/style_notes.md
  02_reference/reference_index.md
  02_reference/literature/
  02_reference/target_journal/
  02_reference/writing_samples/
  03_output/README.md
  03_output/manuscript/
  03_output/tables/
  03_output/figures/
  03_output/supplement/
  03_output/revision/change_log.md
  03_output/revision/local_revision_tasks.md
  03_output/translation/
  03_output/submission/

Do not put project-specific drafts, data, outputs, revision records, or submission packages in shared/.

Chinese mode uses translated folder and template names, including 项目/, 共享/, 01_草稿/, 02_参考/写作样本/, and 03_输出/.

Workflow

1. Confirm the workspace language. Use English if the user does not choose.
2. Identify whether the task is workspace initialization or new-project scaffolding. If the user only asks for base setup, create the workspace roots without a project folder.
3. If creating a project and multiple naming choices are possible, choose a short lowercase slug from the title. Use paper_XX_short_slug when the user has not specified a naming convention.
4. Run the script:

python shared/skills/papersmart-init/scripts/init_papersmart.py --workspace . --language en --project-slug paper_01_short_slug --project-title "Working title"

Add --install-skills when the user asks for dependency installation and network access is allowed:

python shared/skills/papersmart-init/scripts/init_papersmart.py --workspace . --language zh --install-skills

If the user's assistant stores skills in a known custom location, pass it explicitly:

python shared/skills/papersmart-init/scripts/init_papersmart.py --workspace . --language en --install-skills --skills-dir "<skills-dir>"

5. Confirm that the script did not overwrite original user materials. Existing files should be preserved unless the user explicitly requested replacement.
6. If companion skill installation fails, write the exact manual commands to the workspace memory folder.

Companion Skills

PaperSmart works best with these companion skills:

• find-skills, used to choose the smallest useful skill set for a manuscript task.
• humanizer, used when the user wants writing to sound less machine-generated.
• Nature-family academic skills, especially nature-academic-search, nature-writing, nature-polishing, nature-figure, nature-citation, nature-response, nature-data, nature-reader, and nature-paper2ppt.

Installation sources used by the script:

• find-skills: https://github.com/vercel-labs/skills/tree/main/skills/find-skills
• humanizer: https://github.com/blader/humanizer
• Nature skills: https://github.com/Yuan1z0825/nature-skills

The script first checks whether those skill folders already exist under the local assistant skills directory. Skip installation for skills that are already present.

Project Templates

Use precise TODO: markers instead of guessing missing details. Template files should capture:

• Working title, article type, target journal, language, authors, study objective, research questions, methods, data sources, ethics, funding, conflicts, acknowledgements, and missing information in config/project_config.md.
• Source materials, known datasets, table/figure inventories, known data-quality issues, and expected comparisons in 01_draft/data_inventory.md.
• Journal format, writing style, heading style, abstract length, citation style, table/figure rules, and submission constraints in 02_reference/style_notes.md or the Chinese equivalent.
• Literature anchors, target-journal files, writing samples, PDFs, citation files, and "style only, do not cite" notes in 02_reference/reference_index.md or the Chinese equivalent.

Handoff

After initialization, tell the user:

• The workspace or project path created.
• Which companion skills are installed, already present, or still missing.
• Where to place original materials: 01_draft/01_草稿 for user materials and 02_reference/02_参考 for style and literature context.
• That generated outputs must go to 03_output/03_输出.
• To use PaperSmart-new-project for future papers instead of rerunning full initialization.