meeting-transcript
Skill0 starsChanged 9 months agoMITView on GitHub
Two-stage meeting transcript pipeline — ASR cleaning + analytical briefing via Gemini
- Reads credentials
- Installs packages
1 other project uses a version of this.
Adopted by
Most recently taken up by
- huytieu/COG-second-braintheir own version, first seen Jul 30, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- huytieu/COG-second-brain★ 1.3ksame name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository33 other files it gives its agents.
- 13fSkill
- info-lake-healthSkill
- calendarSkill
- chatgpt-organizerSkill
- chatgptSkill
- exploreSkill
- flashbackSkill
- GANSkill
- kbSkill
- learnSkill
- linkSkill
- meeting-backtestSkill
- mocSkill
- notebooklmSkill
- organizer-transcriptSkill
- planSkill
- podcastSkill
- podwise-downloaderSkill
- portfolioSkill
- prompt-optimizerSkill
- researchSkill
- reviewSkill
- socratic-writerSkill
- substackSkill
- supply-chainSkill
- telegramSkill
- thesisSkill
- tradeSkill
- vault-healthSkill
- wechat-haoSkill
- x-bookmarksSkill
- xueqiuSkill
- youtube-downloaderSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 150
After Stage 1 completes, suggest: "Processed N transcripts.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Meeting Transcript Pipeline
Two-stage pipeline for weekly investment meeting transcripts:
- Stage 1 (ASR Clean): Chunks + corrects raw transcripts → verbatim cleaned record
- Stage 2 (Briefing): Full-context analytical summary → per-company investment briefings with action hints
When to Use This Skill
- User says
/meeting-transcriptor asks to clean/process meeting transcripts - User has raw transcript files (txt, docx, pdf) that need ASR error correction
- User wants to batch process multiple meeting recordings into readable notes
Key Features
Stage 1 (ASR Clean):
- Dialect auto-detection: Sichuan dialect vs Mandarin — adjusts prompt automatically
- Smart chunking: Splits long files at speaker boundaries (~15K chars) to avoid Gemini output truncation
- Iterating dictionary: Accumulates correction patterns across runs
- Skip-if-exists: Safely re-runnable — skips already processed files (>10KB)
- Multi-format: Reads .txt, .docx, .pdf source files
- Retry mode: Auto-deletes and reprocesses files below retention threshold
Stage 2 (Briefing):
- Full-context analysis: Sends entire cleaned transcript in one Gemini call (no chunking)
- Per-company structure: 4 anchors per company — 核心观点摘要、正文复述、潜在行动提示、关键跟踪点
- Summary table: One-line IC-ready table for all companies discussed
- Obsidian backlink: Frontmatter links briefing back to cleaned transcript via
[[wikilink]]
Key Files
scripts/batch_process.py— Main processing script (both stages)prompts/cleaning_prompt.md— Stage 1: ASR cleaning prompt (editable)prompts/briefing_prompt.md— Stage 2: Analytical briefing prompt (editable)data/dictionary.json— Accumulated correction dictionary
Dependencies
pip install google-genai python-dotenv pdfplumber python-docxRequires GEMINI_API_KEY in ~/13F-CLAUDE/.env.
Commands
Process all transcripts (default directories)
python ~/.claude/skills/meeting-transcript/scripts/batch_process.pySource: ~/Downloads/周会转写/原文转写/
Output: ~/Documents/会议实录/
Process with custom directories
python ~/.claude/skills/meeting-transcript/scripts/batch_process.py \
--source "~/Downloads/my-transcripts/" \
--output "~/Documents/Obsidian Vault/Meeting Notes/"Process a subset (by index range)
python ~/.claude/skills/meeting-transcript/scripts/batch_process.py --start 10 --end 20Retry files below threshold
# Delete and reprocess any file below 65% retention
python ~/.claude/skills/meeting-transcript/scripts/batch_process.py --retry --threshold 65Use a different Gemini model
python ~/.claude/skills/meeting-transcript/scripts/batch_process.py --model gemini-2.5-proStage 2: Generate analytical briefings
# Generate briefings for all cleaned transcripts
python ~/.claude/skills/meeting-transcript/scripts/batch_process.py --briefing
# Generate briefing for a specific file (by index)
python ~/.claude/skills/meeting-transcript/scripts/batch_process.py --briefing --start 49 --end 50
# Force regeneration (overwrite existing)
python ~/.claude/skills/meeting-transcript/scripts/batch_process.py --briefing --force
# Use a different model for briefing
python ~/.claude/skills/meeting-transcript/scripts/batch_process.py --briefing --briefing-model gemini-3-pro-previewOutput
Stage 1:
- Location:
~/Documents/会议实录/(default) - Naming:
YYYY-MM-DD-会议实录.md - Format: Cleaned markdown with speaker labels, company names bolded, metadata header
Stage 2:
- Location:
~/Documents/会议实录/(same directory) - Naming:
YYYY-MM-DD-周会分析.md - Format: Per-company investment briefing with 4 anchors + IC summary table
- Frontmatter: YAML with tickers, source backlink to cleaned transcript
Workflow
Stage 1: ASR Clean
Source files (txt/docx/pdf)
→ Extract text (pdfplumber/python-docx)
→ Detect dialect (Sichuan markers count)
→ Split into ~15K char chunks at speaker boundaries
→ Send each chunk to Gemini with cleaning prompt + dictionary
→ Concatenate chunks
→ Save as YYYY-MM-DD-会议实录.md
→ Report retention ratio (target ≥80%)
Stage 2: Analytical Briefing
Cleaned transcript (YYYY-MM-DD-会议实录.md)
→ Strip frontmatter
→ Send FULL text to Gemini 2.5 Pro (no chunking)
→ Per-company analysis with 4 anchors
→ Detect tickers, generate frontmatter with source backlink
→ Save as YYYY-MM-DD-周会分析.mdKnown Issues
- Gemini chunk shrinkage: ~20-30% of chunks randomly return ~4K output from ~15K input (API-side issue). Retry usually fixes it.
- Scanned PDFs: pdfplumber can't extract text from image-only PDFs. These will be skipped.
- Rate limiting: 3s between chunks, 5s between files to avoid Gemini 429/500 errors.
Customization
Editing the cleaning prompt
The prompt at prompts/cleaning_prompt.md contains:
- Role definition (TMT/consumer analyst + dialect expert)
- Company name correction dictionary (core mappings)
- Formatting rules (metadata header, speaker labels, filler cleanup)
- Output requirements (≥80% retention, completeness priority)
Edit freely to add new company names, change formatting rules, or adjust the role.
Post-Processing
After Stage 1 completes, suggest: "Processed N transcripts. Run --briefing to generate analytical summaries?"
自动执行: 完成所有处理后,立即对输出文件夹执行 link 扫描(等同于 /link 周会/),为新增笔记添加 [[wikilinks]]。
用户说"跳过 link"时跳过此步。
Growing the dictionary
The data/dictionary.json file accumulates corrections across runs. Add entries manually:
{
"corrections": [
{"from": "错误词", "to": "正确词", "reason": "context note"}
]
}Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

