syndly-setup

name	syndly-setup
description	First-run onboarding. Scaffold ./.syndly/ at the project root and write your brand voice profile plus per-platform tones.

Setup

Walks you through first-run setup of the Syndly skill library on this machine. Run this once per machine before using any other Syndly skill.

What it does

1. Check the existing state. ./.syndly/ is project-level — each repository has its own brand voice. If ./.syndly/brand-voice.md already exists in this project, ask the user: "The folder already exists. Re-run onboarding (overwrites the brand voice profile) or skip?" Only proceed after confirmation.

2. Create the directory tree. Run mkdir -p ./.syndly/platforms ./.syndly/drafts ./.syndly/state from the project root. Confirm each directory was created.

3. Detect the MCP API key. Look for it in env (SYNDLY_API_KEY or SYN_MCP_KEY). If absent, ask the user to paste it and offer to add a hint to .syndly/README.md documenting where to put it. Never persist the key itself to disk under the repo or under .syndly/.

4. Run the brand voice interview. Ask the user, in this order, one question at a time. Wait for the answer before asking the next.

   • "In one sentence, what does this brand sound like? (e.g. measured-expert, witty-peer, warm-mentor)"
   • "Who is the audience? (job titles, expertise level)"
   • "Three topics you want the brand to own."
   • "Two phrases to avoid (anti-patterns)."
   • "The one thing the brand never does."

5. Write .syndly/brand-voice.md. YAML frontmatter for the structured fields plus a freeform notes section beneath. Example:

   ---
   tone: measured-expert
   audience: senior backend engineers running production services
   own_topics:
     - incident retrospectives
     - post-mortem culture
     - platform reliability
   avoid:
     - "synergy"
     - "disrupt"
     - emojis in technical posts
   never: shill for vendor products
   ---
   
   ## Notes
   ...
   

6. Run the per-platform interview. Ask the user, one platform at a time, for a 2-3 sentence tone description:

   • X (Twitter): short, punchy, threads are okay. Voice here may differ from the brand voice default.
   • LinkedIn: longer-form, professional but human, leads with insight or story.
   • Facebook: conversational, image-forward, community-friendly.

   For each platform the user opts to skip, write a .syndly/platforms/<platform>.md containing just a one-line note that the platform is intentionally inactive, so future invocations don't ask again.

7. (Optional) Add the first source. Ask: "Add a first content source now? (URL or RSS feed)" If yes, look up the relevant MCP tool name (start with list_sources to confirm connectivity, then call the add-source tool the server exposes). Always show the user what's about to be added and wait for confirmation before invoking any mutating tool.

8. Print the summary. Confirm:

   • Path to .syndly/
   • Brand voice fields captured
   • Per-platform profiles written / skipped
   • First source added (or skipped)
   • Reminder: invoke /syndly-captures or /syndly-draft next time you want to draft a post.

Things this skill must NOT do

• Write the MCP API key to any file under .syndly/ or anywhere else.
• Invoke mutating MCP tools (anything that creates / updates / removes a row) without showing the user the planned change and waiting for explicit confirmation.
• Skip past the brand voice interview even if the user seems impatient. This document is the foundation the rest of the library reads from.