Ghost

Ghost Skill

Full Ghost Admin API v5 client. HS256 JWT auth and all HTTP calls via stdlib (urllib) - zero external dependencies. Credentials: ~/.openclaw/secrets/ghost_creds · Config: ~/.openclaw/config/ghost/config.json

Trigger phrases

Load this skill immediately when the user says anything like:

• "create / write / draft a blog post / article"
• "publish this post / article", "make this live on Ghost"
• "update / edit [post title or slug]", "change the title of my post"
• "unpublish / revert to draft [post]"
• "create / add a tag", "list my tags on Ghost"
• "create / update / publish a page"
• "upload this image to Ghost"
• "list my posts / drafts", "show me what's in draft"
• "schedule this post for [date]"
• "what's on my Ghost site?", "show site info"

Quick Start

python3 scripts/ghost.py config    # verify credentials + active config
python3 scripts/ghost.py site      # test connection + show site info
python3 scripts/ghost.py posts --limit 3 --fields "id,title,status"

Setup

python3 scripts/setup.py       # interactive: credentials + permissions + connection test
python3 scripts/init.py        # validate all configured permissions against live instance

init.py only runs write/delete tests when allow_delete=true. When allow_delete=false, write tests are skipped - no test artifacts are created, so none can be left behind.

Manual - ~/.openclaw/secrets/ghost_creds (chmod 600):

GHOST_URL=https://your-ghost.example.com
GHOST_ADMIN_KEY=id:secret_hex

Admin API Key: Ghost Admin → Settings → Integrations → Add custom integration → copy Admin API Key.

config.json - behavior restrictions:

Storage & credentials

The skill reads and writes the following paths. All usage is intentional and documented:

Credentials can also be provided via environment variables (GHOST_URL, GHOST_ADMIN_KEY). The skill checks env vars first.

Cleanup on uninstall: clawhub uninstall ghost-admin removes the skill directory. To also remove credentials and config:

python3 scripts/setup.py --cleanup

On reinstall, any existing config at ~/.openclaw/config/ghost/config.json is picked up automatically.

Module usage

from scripts.ghost import GhostClient
gc = GhostClient()
post = gc.create_post("My Title", html="<p>Body</p>", status="draft")
gc.publish_post(post["id"])

CLI reference

# Posts
python3 scripts/ghost.py posts --status published --limit 10
python3 scripts/ghost.py posts --tag devops --fields "id,title,published_at"
python3 scripts/ghost.py post <id_or_slug>
python3 scripts/ghost.py post-create "Title" --html "<p>...</p>" --status draft
python3 scripts/ghost.py post-create "Title" --html-file body.html --tags "DevOps,Linux"
python3 scripts/ghost.py post-create "Title" --html-file body.html --feature-image "https://..."
python3 scripts/ghost.py post-update <id> --fields-json '{"title":"New","custom_excerpt":"..."}'
python3 scripts/ghost.py post-publish <id>
python3 scripts/ghost.py post-unpublish <id>
python3 scripts/ghost.py post-delete <id>          # requires allow_delete: true

# Pages
python3 scripts/ghost.py pages
python3 scripts/ghost.py page-create "About" --html "<p>...</p>"
python3 scripts/ghost.py page-publish <id>

# Tags
python3 scripts/ghost.py tags
python3 scripts/ghost.py tag-create "DevOps" --slug devops --desc "DevOps content"
python3 scripts/ghost.py tag-delete <id>           # requires allow_delete: true

# Images
python3 scripts/ghost.py image-upload ./image.png --alt "Description"

# Members & newsletters (read)
python3 scripts/ghost.py members --limit 20        # requires allow_member_access: true
python3 scripts/ghost.py newsletters
python3 scripts/ghost.py tiers

# Account
python3 scripts/ghost.py site
python3 scripts/ghost.py me
python3 scripts/ghost.py config

Templates

Draft → review → publish

gc = GhostClient()
# 1. Create draft
post = gc.create_post(
    title="My Article",
    html=article_html,
    tags=[{"name": "DevOps"}, {"name": "Linux"}],
    meta_description="Short SEO description",
    status="draft",
)
draft_url = f"{gc.base_url}/ghost/#/editor/post/{post['id']}"
# 2. Return preview link to user for review
# → f"Draft created: {draft_url}"
# 3. After user approval:
gc.publish_post(post["id"])
# → f"Published: {post['url']}"

Research → structured post

# Create a post with full SEO metadata
post = gc.create_post(
    title="Why Rust is Taking Over Systems Programming",
    html=content_html,
    custom_excerpt="A technical deep-dive into Rust's memory model and adoption trends.",
    meta_title="Rust Systems Programming 2025",
    meta_description="Why Rust is replacing C++ in systems programming in 2025.",
    og_title="Rust is Taking Over Systems Programming",
    tags=[{"name": "Rust"}, {"name": "Systems"}],
    feature_image="https://example.com/rust.png",
    status="draft",
)

Content audit by tag

result = gc.list_posts(limit="all", tag="devops", fields="id,title,status,published_at")
posts  = result["posts"]
drafts    = [p for p in posts if p["status"] == "draft"]
published = [p for p in posts if p["status"] == "published"]
# → summarize backlog for user

Batch tag creation

for name in ["DevOps", "Security", "Linux", "Cloud"]:
    try:
        gc.create_tag(name, slug=name.lower())
    except Exception:
        pass  # already exists

Ideas

• Set allow_publish: false + default_status: draft for a "suggest only" mode
• Use default_tags in config for auto-tagging (e.g. always add "AI-assisted")
• Draft from a research summary, share preview link, publish after human review
• List posts --status draft to surface the content backlog for triage
• Upload a feature image first, then embed the returned URL in post HTML

Notes

• updated_at conflict guard: update_post/update_page auto-fetches updated_at if omitted.
• HTML content: Ghost v5 stores Lexical internally but html import works perfectly for agent-generated content.
• allow_publish: false: Status is silently capped to "draft" - no error raised.
• JWT tokens: Generated fresh per request (5-min TTL), no caching needed.
• Slug: Auto-generated from title if omitted. Override with --slug for clean URLs.

Combine with

API reference

See references/api.md for endpoint details, NQL filters, field list, and error codes.

Troubleshooting

See references/troubleshooting.md for common errors and fixes.