/generate-guidelines

MANDATORY FIRST STEP — do this before anything else, including reading sources or processing arguments. Check whether the user has a working folder selected for this session. You must verify this before starting any guideline generation work. If there is no working folder, stop and warn the user: "You don't have a working folder selected. Without one, I can't save guidelines to a file — they'll only exist in this conversation and won't persist to future sessions. Please select a working folder and re-run this command. If you'd like to proceed anyway, let me know." Wait for the user to confirm before continuing.

Generate comprehensive, LLM-ready brand voice guidelines from whatever sources the user provides — brand documents, conversation transcripts, a discovery report from /brand-voice:discover-brand, or direct input.

Process the sources specified in $ARGUMENTS. If none specified, check:

1. Whether a discovery report was generated in this session
2. .claude/brand-voice.local.md for known brand material locations
3. Connected platforms (Notion, Confluence, Google Drive, Box, SharePoint, Gong) for existing materials
4. If nothing is available, suggest running /brand-voice:discover-brand first

Follow the guideline-generation skill instructions to:

1. Identify and classify all available sources (discovery report, documents, transcripts)
2. Delegate to document-analysis and conversation-analysis agents as needed
3. Synthesize findings into unified guidelines with "We Are / We Are Not" table and tone-by-context matrix
4. Assign confidence scores per section
5. Surface open questions with agent recommendations for any ambiguity
6. Present key findings and offer next steps
7. Save guidelines to .claude/brand-voice-guidelines.md inside the user's working folder (archiving any existing file first). Do NOT use a relative path from the agent's current working directory — in Cowork, the agent runs from a plugin cache directory, not the user's project.

After generation, guidelines are saved locally so /brand-voice:enforce-voice can automatically find them in future sessions.

Supported document formats: PDF, PowerPoint, Word, Markdown, plain text. Supported transcript sources: Gong (MCP), Granola (MCP), Notion meeting notes, manual uploads.