peng-post-to-xhs

name	peng-post-to-xhs
description	Generates Xiaohongshu image cards via Google Gemini and publishes them to 小红书. Single Gemini provider with custom GOOGLE_BASE_URL support. Combined image generation + publishing workflow. Use when user asks to create XHS cards, generate 小红书 images, or post to 小红书.

Post to Xiaohongshu

Generate Xiaohongshu image cards from prompts using Google Gemini image models, then publish them to 小红书 — all in one workflow.

Script Directory

{baseDir} = this SKILL.md's directory. Resolve ${BUN_X}: prefer bun; else npx -y bun; else suggest brew install oven-sh/bun/bun.

Script	Purpose
scripts/main.ts	Generate image(s) via Gemini — single or batch
scripts/publish.ts	Full workflow: generate images + publish to 小红书

XHS MCP Scripts (scripts/xhs/)

Wraps xiaohongshu-mcp for 小红书 publish.

Script	Purpose
xhs/mcp-call.sh	Universal MCP tool caller
xhs/start-mcp.sh	Start MCP service
xhs/stop-mcp.sh	Stop MCP service
xhs/status.sh	Check login status
xhs/install-check.sh	Check dependencies (xiaohongshu-mcp, jq)

Prerequisites

1. xiaohongshu-mcp binary: install from GitHub Releases to ~/.local/bin/
2. jq: brew install jq
3. GOOGLE_API_KEY env var set for image generation

Preferences (EXTEND.md)

Check these paths in order; first hit wins:

Path	Scope
.peng-skills/peng-post-to-xhs/EXTEND.md	Project
${XDG_CONFIG_HOME:-$HOME/.config}/peng-skills/peng-post-to-xhs/EXTEND.md	XDG
$HOME/.peng-skills/peng-post-to-xhs/EXTEND.md	User home

Found -> read, parse, apply. Not found -> use defaults.

Supported keys:

Key	Default	Description
default_model	gemini-3.1-flash-image-preview	Gemini model ID
default_quality	2k	normal or 2k
default_aspect_ratio	9:16	Aspect ratio (9:16 for XHS portrait cards)
google_base_url	(Google default)	Custom base URL for proxy (e.g., https://openclaw.chatgo.best)

Value priority: CLI args -> EXTEND.md -> env vars -> defaults.

Environment Variables

Variable	Description
GOOGLE_API_KEY	Google API key (required for image generation)
GEMINI_API_KEY	Alias for GOOGLE_API_KEY
GOOGLE_BASE_URL	Override base URL (same as EXTEND.md google_base_url)
MCP_URL	MCP endpoint (default http://localhost:18060/mcp)

Usage

Image Generation Only

# Single image
${BUN_X} {baseDir}/scripts/main.ts --prompt "A cute cat" --image out.png

# With aspect ratio and quality
${BUN_X} {baseDir}/scripts/main.ts --prompt "XHS cover card" --image cover.png --ar 9:16 --quality 2k

# Prompt from files (concatenated)
${BUN_X} {baseDir}/scripts/main.ts --promptfiles style.md content.md --image out.png

# Batch mode from a directory of prompt files
${BUN_X} {baseDir}/scripts/main.ts --batchdir prompts/ --images-dir output/

Generate + Publish (recommended)

# Full workflow: generate images from prompts then publish to 小红书
${BUN_X} {baseDir}/scripts/publish.ts \
  --batchdir prompts/ \
  --title "我的小红书标题" \
  --content "正文内容，最多1000字" \
  --tags "标签1" "标签2"

# Skip image generation, publish existing images
${BUN_X} {baseDir}/scripts/publish.ts \
  --skip-gen \
  --images output/01-cover.png output/02-content.png \
  --title "标题" \
  --content "正文" \
  --tags "AI工具" "效率"

# Publish with content from a markdown file
${BUN_X} {baseDir}/scripts/publish.ts \
  --images cover.png content.png \
  --title "标题" \
  --content-file article.md \
  --tags "教程" "小红书"

# Schedule publish (1 hour ~ 14 days later)
${BUN_X} {baseDir}/scripts/publish.ts \
  --batchdir prompts/ \
  --title "标题" \
  --content "正文" \
  --schedule-at "2026-06-02T10:00:00+08:00"

First-time Login

The publish script checks login automatically. If not logged in, it prompts you to scan a QR code:

# Manual login (if needed)
cd {baseDir}/scripts/xhs && ./start-mcp.sh
cd {baseDir}/scripts/xhs && ./mcp-call.sh get_login_qrcode
# → scan QR with 小红书 app

Batch Mode

Place .md prompt files in a directory. Each file becomes one image card.

prompts/
├── 01-cover.md
├── 02-content-1.md
├── 03-content-2.md
├── 04-content-3.md
├── 05-content-4.md
└── 06-ending.md

Generate images:

${BUN_X} {baseDir}/scripts/main.ts --batchdir prompts/ --images-dir output/

Or generate and publish in one step:

${BUN_X} {baseDir}/scripts/publish.ts \
  --batchdir prompts/ \
  --title "标题" \
  --content "正文" \
  --tags "aweskill" "AI Agent"

Publish Options

Option	Description
--title <text>	Post title (max 20 chars)
--content <text>	Post body (max 1000 chars)
--content-file <path>	Read body from file
--images <paths...>	Image file paths
--tags <tags...>	Topic tags
--batchdir <dir>	Generate images from prompt dir, then publish
--images-dir <dir>	Output dir for generated images (with --batchdir)
--schedule-at <ISO8601>	Schedule publish (1h~14d)
--skip-gen	Skip generation, publish existing images only
--json	JSON output

小红书 Restrictions

• Title: max 20 characters
• Body: max 1000 characters
• Images: at least 1, formats PNG/JPG
• Tags: topic labels, optional
• Daily limit: 50 posts
• Schedule: 1 hour to 14 days in advance (ISO8601 format)