scribe-workflow

name	scribe-workflow
description	Use when a CoP scribe needs to process a Community of Practice meeting transcript after a session. Orchestrates the full workflow from cleaning the transcript through PR creation, delegating to specialized sub-skills.

Scribe Workflow

Orchestrates the complete post-session workflow for CoP scribes, from transcript cleaning through GitHub PR creation.

When to Use

Invoke this skill when:

• A CoP scribe has a raw transcript from Gemini Live that needs processing
• The scribe needs to archive a cleaned transcript in the a11y-cop repository
• The full end-to-end workflow is needed (not just one isolated step)

Core Pattern

This skill delegates to specialized sub-skills:

1. /clean-session-transcript — Cleans the raw transcript (ASR mishears, name normalization)
2. /extract-meeting-insights — Creates Slack-ready summary in .tmp/session-summary.md
3. /skill-retrospective — Captures scribe feedback for workflow improvement
4. /using-git — Handles proper git/gh operations for PR creation

Workflow Steps

1. Locate transcript: Guide scribe to Google Calendar entry for the CoP Working Session — transcript file is linked there ~5 minutes after meeting ends
2. Prerequisites check: Verify a11y-cop repo exists, sessions/ directory exists (create if needed)
3. Clean transcript: Invoke /clean-session-transcript — handles both export-to-markdown and Gemini find/replace paths
4. Extract insights: Invoke /extract-meeting-insights — generates .tmp/session-summary.md for Slack posting
5. Retrospective: Invoke /skill-retrospective — captures friction points and workflow improvements
6. Create PR: Invoke /using-git — handles branch creation, commit, push, and PR with proper conventions

Scribe Interaction Points

Confirmation required:

• Path selection (A: Export as Markdown, B: Gemini Find/Replace)
• Transcript cleaning approval before archival
• Retrospective participation (optional, can skip)
• Git workflow preference (automated vs manual)

Always ask permission before:

• Executing git operations
• Editing skill files (if retrospective identifies improvements)

File Locations

• Cleaned transcript: a11y-cop/sessions/<YYYY-MM-DD>-session-transcript.md
• Slack summary: .tmp/session-summary.md (gitignored, ready to copy/paste)
• Skills directory: .claude/skills/ (for any retrospective-driven edits)

Critical Rules

• NEVER commit directly to main (always branch + PR)
• ALWAYS get scribe confirmation before major steps
• ALWAYS request permission before executing git operations
• File naming: <YYYY-MM-DD>-session-transcript.md
• Branch naming: transcript/<YYYY-MM-DD>-cop-session
• Commit format: docs(sessions): add cleaned transcript for <date> CoP session

Quick Reference

Typical session duration: 10-15 minutes end-to-end

Path recommendation:

• First-time scribe → Path B (guided Gemini prompts)
• Experienced scribe → Path A (faster, more control, higher Claude token usage)
• Simple transcript → Path B (lower Claude token usage)
• Complex transcript → Path A

Git automation recommendation:

• First-time or unfamiliar with git → Manual (see commands before they run)
• Experienced with git → Automated (faster, single confirmation)

Example Invocation

/scribe-workflow

I have the raw transcript from today's CoP session open in Google Drive and need to clean it, create a Slack summary, and archive it in the repo.

Related Skills

• /clean-session-transcript — Transcript cleaning (ASR corrections, name normalization)
• /extract-meeting-insights — Slack summary generation
• /skill-retrospective — Workflow improvement feedback capture
• /using-git — Proper git/gh interaction guidance for this repo