Setup

---
name: setup
description: >
  This skill should be used when the user asks to create a writing profile,
  set up their blog voice, create a new blog profile, or edit an existing
  profile. Trigger phrases: "draftloom setup", "create a blog profile",
  "new writing profile", "set up my blog voice", "edit my writing profile",
  "draftloom new profile", "/draftloom:setup". Use this skill whenever the
  user wants to configure how Draftloom writes — even if they don't use the
  exact command name.
license: MIT
metadata:
  version: "0.1.0"
  author: Georgios Pilitsoglou
---

# Draftloom Setup Skill

Guides the user through creating or editing a named voice profile for blog post drafting.

## Entry point

On invocation, check `.draftloom/profiles/` for existing profile files.

### If no profiles directory or directory is empty

First, check if `.draftloom/config.json` exists.

If it does not exist, ask: "Store profiles in this project only, or globally? (project/global)"

Wait for the answer. Create `.draftloom/config.json` with the chosen `storage_mode`:
- "project" → `{ "storage_mode": "project", "storage_path": ".draftloom", "version": "0.1.0", "created_at": "<ISO-8601 now>" }`
- "global" → `{ "storage_mode": "global", "storage_path": "~/.draftloom", "version": "0.1.0", "created_at": "<ISO-8601 now>" }`

Then proceed to the "Create mode" section below.

### If profiles exist

Present three options:
1. Create a new profile
2. Edit an existing profile (list profile names)
3. Delete a profile (list profile names, ask for confirmation)

Wait for the user's choice before proceeding.

---

## Edit mode (`/draftloom:setup edit {name}` or user chooses option 2)

Read `.draftloom/config.json` to determine storage_mode. Load the profile JSON from:
- `storage_mode: "global"` → `~/.draftloom/profiles/{name}.json`
- `storage_mode: "project"` (or absent) → `.draftloom/profiles/{name}.json`

Show current values for all fields. Ask the user which field(s) to update. Apply each change one at a time, confirming with the before/after delta format: `Before: {old_value} → After: {new_value}`.

After each confirmed change, ask: "Edit another field? (y/n)" Loop back to field selection if yes.

Save the updated profile to the same path that was loaded. Update `updated_at` to the current ISO-8601 timestamp. Confirm: "Profile '{name}' updated."

Load `references/profile-schema.md` to validate field formats before saving.

---

## Create mode (no profiles exist, or user chooses option 1)

Ask the following 3 essential questions **one at a time**. Wait for the answer to each before asking the next.

Load `references/interview-questions.md` for the exact question wording and validation rules.

### Question 1 — Profile name

Ask for a profile name (slug format, e.g. `george-personal`). Validate:
- Lowercase only, hyphens allowed, no spaces, no special characters
- 3–40 characters
- Must not match an existing profile name in the storage path

Check existing profiles in the storage path. If a profile with the same name already exists, explain the conflict: "A profile named '{name}' already exists. Please choose a different name." Re-ask until a unique valid name is given.

### Question 2 — Target audience

Ask: "Who is your target reader? (e.g. indie hackers, senior engineers, marketing managers)"

Accept free text. Minimum 10 characters — if shorter, ask them to be more specific.

### Question 3 — Tone

Ask: "How would you describe your writing tone? Choose 3–5 adjectives, or pick from these presets:
- authoritative
- conversational
- technical
- witty
- direct
- inspirational

You can mix presets and your own words."

Accept a list of 3–5 adjectives. If fewer than 3 are given, explain: "The minimum is 3 tone adjectives. You've given {N} — please add {3-N} more." Re-ask until at least 3 are provided.

---

## Save profile

Construct the profile JSON using the 3 collected answers. Include only the fields collected (id, audience, tone, storage, created_at, updated_at). Omit all optional fields that were not collected — do not write them as null. Load `references/profile-schema.md` for the full schema.

Determine storage path from `.draftloom/config.json` → `storage_mode`:
- `"global"` → write to `~/.draftloom/profiles/{name}.json`
- `"project"` (or any other value) → write to `.draftloom/profiles/{name}.json`

Load `references/storage-guide.md` if the user asks about storage options.

Write the profile JSON. Confirm using the path that matches the resolved `storage_mode` (the same location you wrote to above):
- `"global"` → "Profile '{name}' saved to `~/.draftloom/profiles/{name}.json`."
- `"project"` (or any other value) → "Profile '{name}' saved to `.draftloom/profiles/{name}.json`."

---

## Post-save: deferred fields hint

After confirming save, show:

```
Profile created. Optional extras you can add later with `/draftloom:setup edit {name}`:
  • Blog URL
  • Content pillars (topic clusters)
  • Distribution channels
  • Typical post length preference
  • Writing inspiration (authors, publications)
  • CTA goal (newsletter, follow, contact)
```

---

## Delete mode (user chooses option 3)

List profiles. Ask which to delete. Ask for confirmation: "Delete '{name}'? This cannot be undone. (y/n)". On confirmation, remove the file. Confirm deletion.

---

## Reference files

Load these files as needed — do not load all at once:

- **`references/interview-questions.md`** — exact wording for all 9 interview questions (3 essential + 6 deferred), validation rules, and accepted input formats
- **`references/profile-schema.md`** — full JSON schema for profile files, field definitions, validation rules, and the distinction between required and optional fields
- **`references/storage-guide.md`** — project vs global storage modes, config.json structure, path resolution, and mode-switching instructions