internal-doc-polisher
Transform raw or transcript-like text into a polished Markdown document for internal sharing. Use when the user provides a text file (any mix of zh_tw, zh_cn, or en) and wants sentence repair, structured headings, concise paragraphs, a 3–7 bullet summary, and an Action Items section when tasks are mentioned.
When & Why to Use This Skill
The Internal Doc Polisher is a specialized Claude skill designed to transform raw, unstructured text or meeting transcripts into professional, well-structured Markdown documents. It solves the problem of messy internal communication by repairing grammar, organizing content with clear headings, and automatically generating executive summaries and action items, ensuring clarity and professional standards for internal sharing.
Use Cases
- Meeting Minutes Automation: Convert messy, verbatim meeting transcripts into structured minutes with clear headings and a dedicated action items section.
- Multilingual Note Refinement: Polishing rough notes that mix English, Simplified Chinese, and Traditional Chinese into a cohesive, professionally formatted internal report.
- Voice-to-Text Cleanup: Refining voice-dictated brainstorms by removing stutters and filler words while preserving technical terminology and core intent.
- Documentation Standardization: Transforming fragmented project updates or 'brain dumps' into standardized Markdown files ready for internal knowledge bases or repositories.
| name | internal-doc-polisher |
|---|---|
| description | Transform raw or transcript-like text into a polished Markdown document for internal sharing. Use when the user provides a text file (any mix of zh_tw, zh_cn, or en) and wants sentence repair, structured headings, concise paragraphs, a 3–7 bullet summary, and an Action Items section when tasks are mentioned. |
Internal Doc Polisher
Overview
Turn unstructured text into a clean Markdown document with repaired sentences, clear sections, and an executive summary. Preserve meaning, keep language consistent with the source, and output a .md file.
Workflow
1) Intake
- Ask for the input file path if not provided.
- Accept any text file containing zh_tw, zh_cn, or en (mixed language is fine).
- Ask for an output path if the user specifies one; otherwise default to
<input_basename>.polished.mdin the same directory.
2) Repair And Clean
- Fix fragments, grammar, and punctuation without changing meaning.
- Remove obvious speech artifacts (stutters, repeated fillers) when they do not change intent.
- Keep proper nouns, numbers, and domain terms intact.
- Normalize spacing rules for the dominant language:
- zh: no extra spaces between Chinese characters; keep English/number tokens spaced.
- en: standard English spacing and punctuation.
3) Restructure
- Create clear headings (
##,###) that match the content flow. - Group related content into concise paragraphs.
- Convert list-like text into bullets.
- Keep the document in the same language as the source unless the user requests translation.
4) Add Summary
- Add a summary section at the beginning with 3–7 bullets.
- Cover key points, outcomes, risks, and action items.
- Keep bullets short and specific.
5) Action Items
- If tasks or next steps are mentioned, add an
## Action Itemssection. - Convert tasks into bullet points; keep owners/dates if present.
6) Output
- Write the final result to the requested
.mdfile. - Confirm the output path in the response.
Output Template
Use this structure, adjusting headings to fit the content:
- Summary bullet 1
- Summary bullet 2
- Summary bullet 3
## Section Title
Concise paragraph.
### Subsection Title
- Bullet
- Bullet
## Action Items
- Task 1
- Task 2