indranilbanerjee/neels-plugins / hermes/digital-marketing-pro/content/skills/cf-connect
cf-connect
VerifiedConnect a registry MCP connector (Webflow, WordPress, Notion) with endpoint and credential steps. "connect WordPress"
/contentforge:cf-connect — Guided Connector Setup
Script location. If your host does not set
${CLAUDE_PLUGIN_ROOT}, the scripts are in this plugin'sscripts/folder, next toskills/. Run them with Python 3.10 or newer:python, orpython3wherepythonis not found (macOS, many Linux systems).
Set up a specific MCP integration for ContentForge with step-by-step instructions tailored to the connector's transport type. Handles HTTP connectors (OAuth-based), npx connectors (API keys, environment variables, .mcp.json entry), unknown connector names (fuzzy matching), and post-setup verification.
The ground truth about connectors
ContentForge ships no .mcp.json (it is gitignored). This is deliberate (v3.9.0 Cowork-safety decision): no connector auto-connects when the plugin is installed. Every connector is opt-in — the user adds it to their own .mcp.json (or connects it at the platform level in Cowork/Desktop Settings → Integrations).
Never claim a connector is pre-wired or already connected. All status claims must come from running python "${CLAUDE_PLUGIN_ROOT}/scripts/connector-status.py" — never from this file or from memory.
When to Use
Use /contentforge:cf-connect <name> when:
- You want to connect a specific service (e.g.,
/contentforge:cf-connect wordpress) - A skill told you a connector is missing and suggested using this command
- You're setting up a new ContentForge installation and configuring integrations
- You need the exact environment variables and
.mcp.jsonentry for an npx connector - You want to verify an existing connector is working correctly
What This Command Does
- Look Up Connector — Find the connector in the registry, handle typos and close matches
- Check Current Status — Determine if already configured or needs setup
- Present Setup Instructions — Tailored to transport type (HTTP vs npx)
- Guide Credential Acquisition — Where to get API keys, which permissions to set
- Provide .mcp.json Entry — Ready-to-paste configuration
- Verify After Setup — Confirm environment variables are set and configuration is valid
Required Inputs
Required:
- Connector name — The service to connect (e.g.,
notion,wordpress,google-sheets,canva) - Fuzzy matching is supported:
wpmatcheswordpress,gsheetsmatchesgoogle-sheets,gamatchesgoogle-analytics
Optional:
- Environment —
coworkorclaude-code(auto-detected if not specified)
What Happens
Step 1: Look Up Connector (Instant)
Run the setup guide lookup:
python "${CLAUDE_PLUGIN_ROOT}/scripts/connector-status.py" --action setup-guide --name <connector>
This returns:
- Connector metadata (name, category, description, transport type)
- Whether it's already configured in this installation
- Setup steps specific to the transport type
- Skills unlocked by this connector
- For npx: required environment variables and
.mcp.jsonentry
If the connector name is not found, check common aliases before showing an error:
| Input | Matches To |
|---|---|
wp, wordpress | wordpress |
gsheets, sheets, google-sheets | google-sheets |
gdrive, drive, google-drive | google-drive |
ga, ga4, google-analytics | google-analytics |
gsc, search-console | google-search-console |
gcal, calendar | google-calendar |
x, twitter | twitter-x |
li, linkedin | linkedin-publishing |
ig, insta | instagram |
hs, hubspot | hubspot-cms |
sarvam | sarvam-ai |
If no match is found after alias lookup, run python "${CLAUDE_PLUGIN_ROOT}/scripts/connector-status.py" --action list-available and present the actual list.
Step 2: Check Current Status (Instant)
python "${CLAUDE_PLUGIN_ROOT}/scripts/connector-status.py" --action check --name <connector>
This returns connected or not_connected, plus (for npx connectors) which env vars are set vs missing. Render exactly what the script reports.
If already configured: confirm to the user, list the skills it enables (from the script JSON), and stop — no action needed.
If not configured, proceed to Step 3.
Step 3: Present Setup Instructions
Instructions differ based on transport type. Both types require the user to add an entry to .mcp.json (or connect at the platform level) — nothing is pre-wired.
HTTP Connectors (e.g., Notion, Canva, Figma, Webflow, Slack, Gmail, Google Calendar, Ahrefs)
HTTP connectors are the easiest: one .mcp.json entry, no API keys for OAuth-based ones, and they work in both Cowork and Claude Code.
Setup pattern to walk the user through:
- Get the verified endpoint URL. Use the endpoint from the setup-guide script output, cross-checked against
.mcp.json.connectors-reference(the plugin's catalog of verified HTTP endpoints). Do not guess endpoint URLs from memory — unverified URLs waste the user's time. - Add the entry to
.mcp.jsonin the plugin root:
In Cowork/Claude Desktop, the platform-level integration (Settings → Integrations) may be simpler — mention it when available."notion": { "type": "http", "url": "<verified endpoint from the reference file>" } - Restart the session so the new entry is picked up.
- Authorize on first use. The first time a skill accesses the connector, the platform shows an OAuth prompt. Sign in and grant permissions. The token is managed by the platform afterward.
npx Connectors (vendor-published packages only)
ContentForge gives a runnable npx command only for a package the service's vendor publishes, and the list is config/verified-npx-packages.json (the package name, the npm account expected on it, the vendor's repository). Today that is DeepL. npx -y <name> downloads and runs whatever is published under that name with the user's API keys in its environment, so a community package, or a name nobody has registered, is never offered.
Every other connector whose setup-guide says "transport": "manual" has no verified connector. For those, relay the guide's options and stop there: the vendor's own MCP server if it publishes one (an HTTP address from the vendor's documentation), an aggregator entry from .mcp.json.connectors-reference, or skip it. Do not search npm for a package, guess a package name, suggest npx -y for it, or write environment variables for it. If the user mentions a community package themselves, say it is an unverified third party and that they should read its source and check its publisher before running it; give no command for it.
For a connector whose guide says "transport": "npx" (a verified package), the pattern is:
- Get the package name and env vars from
python "${CLAUDE_PLUGIN_ROOT}/scripts/connector-status.py" --action setup-guide --name <connector>. - Obtain credentials from the service (for DeepL: an API key from the DeepL account page).
- Set environment variables (shell profile or
.envloaded before Claude Code starts; on Windows: System Properties → Environment Variables, then restart the terminal):export DEEPL_API_KEY="<your DeepL API key>" - Add the
.mcp.jsonentry (use the exact block from the setup-guide output):"deepl": { "command": "npx", "args": ["-y", "deepl-mcp-server@1.3.14"], "env": { "DEEPL_API_KEY": "${DEEPL_API_KEY}" } } - Restart Claude Code and verify (Step 5).
Never hardcode credentials in .mcp.json or plugin files — always reference environment variables.
Step 4: Handle Unknown Connectors
When a connector name is not in the registry and no fuzzy match exists:
- Say clearly it's not in the registry.
- Suggest the closest same-category connectors from
--action list-availableoutput. - Offer manual configuration: an HTTP entry (
"type": "http", "url": ...) if the service publishes an MCP endpoint. Do not offer an npx entry: ContentForge gives a command only for the vendor-published packages inconfig/verified-npx-packages.json. - Note: manually added connectors work with Claude but won't appear in
/contentforge:cf-integrationsuntil added to the registry. - For a fully guided custom flow, hand off to
/contentforge:cf-add-integration.
Step 5: Verify After Setup
For npx connectors, after the user follows the setup steps:
- Re-run
python "${CLAUDE_PLUGIN_ROOT}/scripts/connector-status.py" --action check --name <connector>and render the result (env vars set/missing,.mcp.jsonentry found). - Remind the user to restart Claude Code —
.mcp.jsonis read at startup only. - Suggest a first test, e.g.
/contentforge:cf-translateon a short paragraph for DeepL.
For HTTP connectors, verification happens implicitly: the OAuth prompt appears on first real use.
Output
The complete setup guide includes:
| Section | Description |
|---|---|
| Connector Info | Name, category, description, transport type (from script) |
| Current Status | Configured or not, with details (from script) |
| Setup Steps | Numbered walkthrough tailored to transport type |
| Skills Unlocked | Which ContentForge skills this connector enables (from script) |
| Credential Requirements | For npx: env vars needed with set/missing status |
| .mcp.json Entry | Ready-to-paste JSON block |
| Environment Notes | Cowork vs Claude Code compatibility, Node.js requirements |
| Verification Steps | How to confirm the connector works after setup |
Troubleshooting
"Unknown connector" even though you typed the right name
- Connector names use lowercase kebab-case:
google-sheets, notGoogle Sheetsorgooglesheets - Try the fuzzy match: the skill checks common aliases automatically
- Run
python "${CLAUDE_PLUGIN_ROOT}/scripts/connector-status.py" --action list-availableto see all valid names
npx connector configured but skills still can't use it
- Restart Claude Code after modifying
.mcp.json— changes are read at startup only - Verify the npm package name (
npm view <pkg> version) - Check that Node.js and npx are in your PATH:
npx --version
Environment variables set but still showing "missing"
- Variables must be exported, not just assigned:
export VAR=value .envfiles must be loaded before the Claude Code session starts- On Windows, set them in System Properties → Environment Variables, then restart the terminal
OAuth prompt never appears for an HTTP connector
- HTTP connectors authenticate only when a skill actively requests data from them
- Confirm the entry exists in
.mcp.jsonand the session was restarted after adding it
"npx: command not found"
- Install Node.js 18+ from https://nodejs.org, then verify
node --versionandnpx --version
Connector works in Claude Code but not in Cowork
- npx connectors are Claude Code only — they require local Node.js
- In Cowork, prefer HTTP connectors or platform-level integrations (Settings → Integrations)
- Run
/contentforge:cf-integrationsto see which configured connectors are Cowork-compatible
Example Workflows
Workflow 1: Set Up WordPress Publishing
1. /contentforge:cf-connect wordpress
-> Follow steps: verify package on npm, get application password, set env vars, add to .mcp.json
2. Restart Claude Code
3. /contentforge:cf-integrations
-> Verify WordPress shows as configured (from script output)
4. /contentforge:create-content "Your Topic" --type=blog --brand=YourBrand
5. /contentforge:cf-publish --platform=wordpress --status=draft
Workflow 2: Add Google Sheets for Batch Processing
1. /contentforge:cf-connect google-sheets
-> Follow steps: create service account, download credentials JSON, set env var, add to .mcp.json
2. Restart Claude Code
3. /contentforge:batch-process <sheet URL>
-> Reads requirements from the sheet, processes them one at a time (checkpointed queue)
Workflow 3: Connect SEO Tools for Data-Driven Briefs
1. /contentforge:cf-connect ahrefs
-> HTTP connector: add the verified endpoint to .mcp.json, restart, authorize on first use
2. /contentforge:cf-brief "AI in Healthcare 2026" --brand=AcmeMed
-> Brief now includes real keyword data from Ahrefs
Optional Connectors (Tier-2 External Validators)
The following connectors are optional and enable enhanced validation passes during specific pipeline phases. They are auto-detected if available (via MCP discovery) and never required — the built-in validators run everywhere without them.
| Connector | Service Type | Use Case |
|---|---|---|
| AI-detection (optional) | any MCP-served detector (e.g. a Hugging Face Space) | Phase 6.5 external validation pass — advisory only; the built-in tell-scan runs everywhere without it |
Agent Used
None. This skill is entirely script-driven using scripts/connector-status.py.
Related Skills
- /contentforge:cf-integrations — Full integration status dashboard
- /contentforge:cf-add-integration — Guided flow for services not in the registry
- /contentforge:create-content — Main content production pipeline
- /contentforge:batch-process — Sequential, checkpointed batch processing
Script: python "${CLAUDE_PLUGIN_ROOT}/scripts/connector-status.py" --action setup-guide --name <connector>
Network Required: No for status checks (reads local config and env vars); yes only for npm view package verification
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 469k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |