Skip to content
Back to search

latex-to-word-workflow

Agent Skill for polished LaTeX-to-Word conversion with live Zotero citations

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/hajimi-kun--latex-to-word-workflow/badge.svg)](/listings/hajimi-kun--latex-to-word-workflow)

HTML

<a href="/listings/hajimi-kun--latex-to-word-workflow"><img src="/api/listings/hajimi-kun--latex-to-word-workflow/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 135 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. # LaTeX to Word Workflow
  2. An agent skill for planning, executing, and reviewing project-specific
  3. academic LaTeX-to-Word conversions.
  4. Current version: **3.0.0**
  5. ## Purpose
  6. LaTeX-to-Word conversion is highly dependent on the manuscript, target
  7. template, citation workflow, numbering semantics, and local Word environment.
  8. This skill therefore provides a lean routing contract, detailed on-demand
  9. guidance, read-only evidence tools, and acceptance gates instead of a fixed
  10. conversion pipeline.
  11. The agent should:
  12. - understand the complete LaTeX project and compile it when the host exposes the
  13. toolchain, otherwise record compilation evidence as pending;
  14. - inspect the target Word requirements and preserve original files;
  15. - start with the smallest reliable approach, normally Pandoc with a reference
  16. DOCX, and add project-local processing only for observed needs;
  17. - prove the LaTeX–Pandoc–Word semantic mapping with representative content;
  18. a reference DOCX does not align differently named styles or create target
  19. behavior that Pandoc does not emit;
  20. - preserve scientific content and numbering semantics ahead of editability;
  21. - verify rather than merely claim live citations;
  22. - probe Zotero capability only through host-exposed evidence and prefer confirmed
  23. live fields unless static citations are explicitly required or probing fails;
  24. - generate and validate native Word cross-references by default, allowing
  25. static references only after an explicit informed user opt-out;
  26. - validate high-risk objects and disclose approved exceptions;
  27. - preserve prose chemical formulae and units as ordinary subscript/superscript
  28. text unless they are genuine mathematical expressions;
  29. - use the user's desktop Word as the authoritative renderer for Word-specific
  30. fields, numbering, and template behavior;
  31. - hand off exact Word/Zotero steps for the user to run, and independently re-audit
  32. the refreshed file when the user returns it.
  33. ## Installable skill
  34. The only installable and released skill directory is
  35. [`latex-to-word-workflow/`](latex-to-word-workflow/).
  36. Repository-level behavior evals are kept in [`tests/`](tests/). They are
  37. versioned with the skill so trigger and workflow expectations can be reviewed,
  38. but are excluded from `git archive` release artifacts.
  39. ```powershell
  40. git clone https://github.com/hajimi-kun/latex-to-word-workflow.git
  41. Copy-Item -Recurse .\latex-to-word-workflow\latex-to-word-workflow "$env:USERPROFILE\.codex\skills\"
  42. ```
  43. ```bash
  44. git clone https://github.com/hajimi-kun/latex-to-word-workflow.git
  45. cp -R latex-to-word-workflow/latex-to-word-workflow ~/.claude/skills/
  46. ```
  47. No Python package or bundled conversion CLI is required. The tools needed for
  48. a conversion depend on the project and may include Pandoc, a LaTeX toolchain,
  49. Microsoft Word, Zotero, or small project-local adapters.
  50. ## Skill layout
  51. ```text
  52. latex-to-word-workflow/
  53. ├── SKILL.md
  54. ├── agents/
  55. │ └── openai.yaml
  56. ├── references/
  57. │ ├── conversion-playbook.md
  58. │ └── template-distillation.md
  59. └── scripts/
  60. ├── docx_audit.py
  61. └── latex_manifest.py
  62. ```
  63. - `SKILL.md` routes the task and contains the default workflow, decision rules,
  64. required checks, and recurring gotchas.
  65. - `conversion-playbook.md` supplies detail only for complex templates,
  66. citations, cross-references, high-risk objects, validation, and handoff.
  67. - `template-distillation.md` explains how to extract paragraph, character,
  68. numbering, object, and transition semantics from a Word target before the
  69. Pandoc probe.
  70. - `latex_manifest.py` generates a deterministic, read-only source inventory.
  71. - `docx_audit.py` checks the OpenXML package, fields, bookmark scopes, applied
  72. styles, unresolved markers, and refresh-save-reopen structural deltas without
  73. modifying the DOCX.
  74. - `openai.yaml` provides Codex UI metadata and a default invocation prompt.
  75. ## Design principles
  76. - **Submission-grade product:** the primary deliverable is a reviewable Word
  77. candidate, not a quick conversion draft.
  78. - **Guidance plus evidence:** conversion choices remain contextual while stable,
  79. read-only checks are scripted.
  80. - **One default:** begin with editable Pandoc output and escalate only when an
  81. observed requirement justifies more processing.
  82. - **Project-local adaptation:** do not turn one manuscript's workaround into a
  83. universal source requirement.
  84. - **Explicit semantic bridge:** inspect what Pandoc actually emits, map it to
  85. target styles, and post-process Word-only behavior deliberately.
  86. - **Honest semantics:** visible citation or numbering text is not proof of a
  87. dynamic Word/Zotero field; native Word cross-references are the default when
  88. the source contains references.
  89. - **Proportionate validation:** inspect high-risk content without requiring
  90. page-by-page LaTeX/PDF equality by default.
  91. - **Capability-neutral execution:** never assume shell, process, GUI, COM, export,
  92. or rendering access. Record unobservable capabilities as unknown and convert
  93. optional desktop work into explicit user-run gates.
  94. - **Word acceptance:** desktop Word validation and user open-and-confirm are
  95. required; package inspection and LibreOffice cannot substitute for them.
  96. ## Repository checks
  97. The repository-level eval definitions are:
  98. - `tests/trigger-cases.md`: three should-trigger and three should-not-trigger
  99. prompts;
  100. - `tests/workflow-cases.md`: template-semantic cases that must be recognized
  101. before full conversion;
  102. - `tests/expected-behaviors.md`: scoring, with/without-skill comparison, and
  103. cross-harness result recording.
  104. Run these prompts in the supported harnesses and record results in a separate
  105. local work log. They evaluate skill behavior; they do not attempt to test a
  106. universal document converter.
  107. ## License
  108. MIT. See [LICENSE](LICENSE).
  109. ## Community
  110. - [LINUX DO](https://linux.do/)

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.