lov-xbti-creator

xbti-creator — Create Your Own BTI Personality Test

Generate a complete, deployable BTI personality test website from a theme and preferences. Based on the proven XBTI architecture (React + Vite), with AI-generated content.

When to Use

User wants to create a themed personality test (e.g., LBTI for lobsters, FBTI for founders)
User says "创建一个XX人格测试" or "make a custom BTI"

Workflow (MANDATORY)

Step 1: Collect Theme Info

Use AskUserQuestion to gather:

Field	Example	Required
BTI name	LBTI	Yes
Theme	Personality analysis of a lobster	Yes
Style preference	Humorous and sarcastic / warm and healing / academic and serious	Yes
Number of personalities	16-25 (default 20)	No
Special easter egg personality	Yes/No (default Yes)	No
Project directory	Default ~/projects/{NAME}	No
Submit to Gallery	Yes/No (default No) — submit to xbti.example.com for others to experience	No

Step 2: Environment Setup

After collecting user input, silently check and auto-fix the environment:

# 1. Node.js (required)
node --version  # Need 18+

# 2. Detect package manager
command -v pnpm && echo "USE_PM=pnpm" || command -v bun && echo "USE_PM=bun" || echo "USE_PM=npm"

# 3. git
git --version

# 4. image-creator skill
npx skills add image-creator -g -y

# 5. Zenmux API key
[ -n "$ZENMUX_API_KEY" ] && echo "OK" || echo "MISSING"

# 6. Python deps
python3 -c "import google.genai; from PIL import Image" 2>/dev/null && echo "OK" || echo "MISSING"

Auto-fix (no user interaction needed for these):

lov-image-creator missing →
```
npx skills add image-creator -g -y
```

Python deps missing →

pip install google-genai Pillow --break-system-packages 2>/dev/null || pip3 install google-genai Pillow

Require user input only if truly blocked:

Node.js missing → tell user to install and stop
ZENMUX_API_KEY missing → ask user to provide it:
1. Register Zenmux: https://zenmux.ai/invite/K6KT2X (invite code, free credits available)
2. Get API Key: https://zenmux.ai/settings/keys
3. Setup: export ZENMUX_API_KEY="your-key" (recommended to add to ~/.zshrc)

Do NOT proceed to Step 3 until environment is fully ready.

Step 3: Design Dimension System

Based on the theme, design 5-dimension model, each with 3 sub-dimensions = 15 dimensions total.

Rules:

Dimension names must be thematic (e.g., for lobster BTI: "钳力模型", "深海适应模型")
Each sub-dimension measures a distinct aspect on a L/M/H scale
Keep dimension IDs short: D1-D15 or thematic prefixes (C1, C2, C3, S1, S2, S3...)
Write L/M/H explanations for each dimension (one-line, matching the BTI's tone)

Output: src/data/dimensions.js — see references/xbti-data-format.md for exact format.

Step 4: Generate Questions

Generate 30 questions (2 per dimension), each with 3 options (value 1, 2, 3).

Rules:

Questions must match the theme and tone
Option values: 1 = low on that dimension, 2 = medium, 3 = high
Mix question styles: direct statements, scenarios, hypotheticals
Include 1-2 memorable/funny questions per model for engagement
Optionally: add 1-2 special trigger questions for a hidden personality type

Output: src/data/questions.js

Step 5: Design Personality Types

Generate 16-25 personality types, each with:

Field	Description
`code`	4-6 char English code (creative, memorable)
`cn`	2-3 char Chinese name
`pattern`	15-char L/M/H pattern matching the dimension system
`intro`	One-line tagline (catchy, in-character)
`desc`	200-400 char personality description (match chosen tone)

Rules:

Codes should be thematic and creative (not generic like TYPE-A)
Patterns must cover the L/M/H space well — avoid clustering
Include 1 fallback type (for < 60% match) and optionally 1 hidden type
Descriptions should be entertaining and shareable — this is what goes viral

Output: src/data/types.js

Step 6: Scaffold the Project

Clone the XBTI template from GitHub and clean up:

git clone https://example.com/skills/XBTI.git {TARGET_DIR}
cd {TARGET_DIR}

# Remove origin-specific files, keep the engine
rm -rf .git .vercel image/* CHANGELOG.md docs/
mkdir -p image
git init

# Fix packageManager field for user's environment
# If user uses npm: remove the packageManager field entirely
# If user uses bun: change to bun
# If user uses pnpm: keep as-is

packageManager handling (important for portability):

The cloned package.json has "packageManager": "pnpm@x.y.z"
If user doesn't use pnpm, remove this field to avoid errors
Then run {PM} install with whatever package manager the user has

Then overwrite the data files with generated content:

src/data/dimensions.js
src/data/questions.js
src/data/types.js

Update branding in:

index.html — title, meta tags
src/components/IntroScreen.jsx — hero title, subtitle, attribution
src/index.css — theme colors (optional, ask user)
package.json — name, description, remove packageManager if needed

Keep src/logic/scoring.js as-is — the algorithm is universal. Only update import paths if dimension IDs changed.

Step 7: Generate Avatar Images

For each personality type, programmatically generate an avatar image using the lov-image-creator skill.

Prompt Crafting (nano-banana-pro style):

For each personality type, craft a prompt optimized for stylized character generation:

[Subject: character visual based on personality], [Action/Pose],
[Environment: thematic background], chibi style character,
cute cartoon illustration, simple flat design, vibrant colors,
white background, centered composition,
best quality, masterpiece, ultra high res, 8k

Examples:

CTRL (Controller): "A confident chibi character with arms crossed, wearing a golden crown, standing on a glowing control panel with buttons and levers, chibi style, cute cartoon illustration, simple flat design, vibrant colors, white background, centered composition, best quality, masterpiece, 8k"
DEAD (The Departed): "A peaceful chibi ghost character floating with closed eyes, translucent body, surrounded by ZZZ symbols and stars, chibi style, cute cartoon illustration, simple flat design, soft muted colors, white background, centered composition, best quality, masterpiece, 8k"

Generation Process:

Craft all prompts (one per personality type), present to user
Ask user: "先生成一个预览确认风格，还是直接批量生成全部？"

Preview mode: Generate one image, show result, iterate on prompt style if needed:

IMAGE_CREATOR_DIR="${SKILL_IMAGE_CREATOR_DIR:?set SKILL_IMAGE_CREATOR_DIR}"
python3 "$IMAGE_CREATOR_DIR/gen_image.py" \
  "PROMPT_HERE" \
  -o {TARGET_DIR}/image/{CODE}.png \
  -q medium --no-open

Batch mode: Once style confirmed, loop through all types:

# For each personality type:
IMAGE_CREATOR_DIR="${SKILL_IMAGE_CREATOR_DIR:?set SKILL_IMAGE_CREATOR_DIR}"
python3 "$IMAGE_CREATOR_DIR/gen_image.py" \
  "PROMPT_FOR_TYPE" \
  -o {TARGET_DIR}/image/{CODE}.png \
  -q medium --no-open

Use Read tool to display each generated image to user for quick review.

Verify all images exist:
```
ls -la {TARGET_DIR}/image/
```

IMPORTANT: All image generation is fully automated via CLI — do NOT ask user to manually download or operate any web UI.

Step 8: Install & Launch

cd {TARGET_DIR}

# Install dependencies with user's package manager
{PM} install

# Start dev server
{PM} run dev
# or: npx vite (universal fallback)

Tell the user the dev server URL (usually http://localhost:5173) and invite them to test the full flow.

Step 9: Submit to XBTI Gallery (Optional)

Skip this step if user chose NOT to submit in Step 1.

If user chose to submit, create a PR to skill-publisher/XBTI repo:

# 1. Fork & clone the XBTI repo
cd /tmp
gh repo fork skill-publisher/XBTI --clone
cd XBTI

# 2. Create case directory (follow existing structure: cases/sbti/, cases/cbti/)
CASE_NAME=$(echo "{BTI_NAME}" | tr '[:upper:]' '[:lower:]')
mkdir -p cases/${CASE_NAME}/images

# 3. Copy data files + images
cp {TARGET_DIR}/src/data/dimensions.js cases/${CASE_NAME}/
cp {TARGET_DIR}/src/data/questions.js cases/${CASE_NAME}/
cp {TARGET_DIR}/src/data/types.js cases/${CASE_NAME}/
cp {TARGET_DIR}/image/*.png cases/${CASE_NAME}/images/

# 4. Create index.js (re-export entry point, required by registry)
cat > cases/${CASE_NAME}/index.js << EOF
export const meta = { id: '${CASE_NAME}', name: '${BTI_NAME}', desc: '${THEME}', author: '$(git config user.name)', authorUrl: '' };
export { dimensionMeta, dimensionOrder, DIM_EXPLANATIONS } from './dimensions.js';
export { questions } from './questions.js';
export { TYPE_LIBRARY, TYPE_IMAGES, NORMAL_TYPES } from './types.js';
EOF

# 5. Create case.json (combined metadata + data for static access)
#    Use Node.js to merge all data into a single JSON file:
node -e "
const dim = require('./cases/${CASE_NAME}/dimensions.js');
const q = require('./cases/${CASE_NAME}/questions.js');
const t = require('./cases/${CASE_NAME}/types.js');
const data = {
  meta: { id: '${CASE_NAME}', name: '${BTI_NAME}', desc: '${THEME}', author: '$(git config user.name)', authorUrl: '' },
  ...dim, ...q, ...t
};
require('fs').writeFileSync('./cases/${CASE_NAME}/case.json', JSON.stringify(data, null, 2));
"

# 6. Update cases/registry.js — add import and entry
#    Read current registry, add new import and array entry
# 7. Update cases/index.json — add new case metadata entry

# 8. Create PR
git checkout -b add-case/${CASE_NAME}
git add cases/${CASE_NAME} cases/registry.js cases/index.json
git commit -m "feat: add ${BTI_NAME} case (${THEME})"
git push origin add-case/${CASE_NAME}
gh pr create \
  --repo skill-publisher/XBTI \
  --title "feat: add ${BTI_NAME} (${THEME})" \
  --body "$(cat <<'PREOF'
## New BTI Case

- **Name**: {BTI_NAME}
- **Theme**: {THEME}
- **Style**: {STYLE}
- **Types**: {TYPE_COUNT} personality types

Generated with [xbti-creator](https://example.com/skills/skills)
PREOF
)"

Update cases/registry.js: Add import * as {CASE_NAME} from './{CASE_NAME}'; and append to the CASES array.

Update cases/index.json: Append {"id":"{CASE_NAME}","name":"{BTI_NAME}","desc":"{THEME}","author":"...","authorUrl":"..."} to the array.

Tell user the PR URL and that it will appear on xbti.example.com after merge.

# Clean up
rm -rf /tmp/XBTI

Reference Files

references/xbti-data-format.md — exact data structure format for all files

User Configuration

Set SKILL_IMAGE_CREATOR_DIR to the installed lov-image-creator skill directory when invoking its gen_image.py script directly. If the runtime can invoke the image-creator skill natively, prefer that path instead of a file path.

Design Principles

Theme consistency — every element (dimensions, questions, types, avatars) must feel thematic
Shareability — personality descriptions should be entertaining enough to screenshot and share
Balance — personality types should cover the spectrum, no type should feel "objectively better"
Humor — even serious themes benefit from wit in the descriptions
Completeness — the output should be a fully working, deployable website

Runtime context (shared)

Before running, read this Skill package's skill.yaml, with skill-runtime/v1 context provided by the host. Field resolution order is: current request, project context, personal Preferences, brand Profile, general defaults.

Use only fields declared in the Manifest; Profile stores public brand facts, Preferences stores personal work preferences.
required: true When a field is missing, ask the user one focused question according to the Manifest's question configuration; save the answer only after the user explicitly agrees.
Errors provide a copyable context_id, field path, and source; diagnostic content avoids secrets, full private paths, and raw configuration.

About this skill

When to Use

Workflow (MANDATORY)

Step 1: Collect Theme Info

Step 2: Environment Setup

Step 3: Design Dimension System

Step 4: Generate Questions

Step 5: Design Personality Types

Step 6: Scaffold the Project

Step 7: Generate Avatar Images

Step 8: Install & Launch

Step 9: Submit to XBTI Gallery (Optional)

Reference Files

User Configuration

Design Principles

Runtime context (shared)