new-client

/new-client

Stand up a new client project at projects/$1/. The slug ($1) is kebab-case and becomes the output filename prefix everywhere. A project is pure data — it inherits every capability from this parent framework (see the root CLAUDE.md). This command does the scaffold, then interviews the user for the profile that the skills downstream depend on.

Steps

1. Guard. If projects/$1/ already exists, stop and tell the user (offer to open client.yml instead). Slug must be kebab-case — lowercase, hyphens, no spaces.

2. Scaffold the folders.

   mkdir -p projects/$1/{inputs,research,deliverables,data}
   

   • inputs/ — briefs, exports, source material the client gave us.
   • research/ — keyword / competitor / GEO / audit outputs we generate.
   • deliverables/ — proposals, reports, comparison pages (the handover).
   • data/ — snapshots, rank history, csv (data/raw/ is gitignored).

3. Copy the templates (don't author from scratch — copy then edit):

   cp templates/client.yml.example projects/$1/client.yml
   cp templates/project/CLAUDE.md   projects/$1/CLAUDE.md
   

4. Interview the user for the essentials. Ask in one batched message; let them answer what they know and leave the rest blank. Capture:

   • Name + domain (apex, e.g. acme.com).
   • Industry (one line) and market (primary geo) + languages.
   • Positioning — 2–3 lines: what they sell, to whom, the wedge.
   • ICP — the ideal customer profiles / segments.
   • Competitors — name + domain for each (used by competitor + comparison skills).
   • Target keywords — seed terms for keyword / rank / programmatic work.
   • Brand colors — primary and accent hex (for proposal/report decks).
   • Ahrefs ids — project_id (Rank Tracker / Site Audit) and brand_radar_report_id (GEO). If the user doesn't have them, offer to list with management-projects and management-brand-radar-reports (Ahrefs MCP) and match by domain, then paste the ids in.
   • Owned analytics (optional) — GA4 property id, GSC site (sc-domain:… or URL).