Hopkin – Paid Ads: Meta, TikTok, Google, LinkedIn, Reddit

Query ad platform data (Meta, Google, LinkedIn, Reddit) using the Hopkin CLI. Use this skill when the user asks about ad accounts, campaigns, ad performance,...

Security Scan

The skill's instructions, required actions, and requested inputs are coherent with a read-only CLI for querying ad platforms; nothing requested is disproportionate, though there are a few minor documentation gaps you should be aware of.

Purpose & Capability

The skill's name and description (querying ad platform data via the Hopkin CLI) match the SKILL.md instructions which show hopkin commands for Meta, Google, LinkedIn, and Reddit. One small mismatch: the top-level name/description mention TikTok, but the SKILL.md never documents hopkin 'tiktok' commands—this is likely an oversight in docs, not a functional red flag.

Instruction Scope

The instructions stay on-topic: install the hopkin CLI, check auth, run hopkin <platform> <resource> [verb] [flags], and prefer --json for parsing. The SKILL.md explicitly limits use to read-only queries and tells the agent to ask the user for an API key if not authenticated. It does not instruct the agent to read unrelated files, environment variables, or exfiltrate data to unexpected endpoints.

Install Mechanism

This is an instruction-only skill (no install spec bundled). The doc tells users/agents to install @hopkin/cli globally via npm. That is a normal, expected install step for a CLI but is executed outside the packaged skill. You may want to verify the npm package and publisher before installing and prefer running installations manually (not pasting install commands into a privileged agent session).

Credentials

The skill does not require or declare environment variables or unrelated credentials. It does instruct the agent to ask the user for the Hopkin API key and run 'hopkin auth set-key <API_KEY>' — this is proportional and necessary for the stated read-only queries. There are no requests for unrelated secrets or cloud credentials.

Persistence & Privilege

No special persistence is requested: always is false, and the skill does not ask to modify other skills or system-wide settings. The ability to authenticate and store a hopkin API key (via the hopkin CLI) is normal for a CLI-based integration and is limited to the Hopkin tool.

Assessment

This skill appears to do what it says: it guides an agent to install and use the Hopkin CLI to read advertising data. Before installing or providing keys: (1) verify the @hopkin/cli npm package and its publisher (npmjs.org) and prefer manual installation in a controlled shell, (2) do not paste API keys directly into untrusted chats—use the hopkin CLI's auth command or a secure secret manager, (3) note the SKILL.md omits TikTok examples despite the name — confirm whether Hopkin supports TikTok if you need it, and (4) review the Hopkin account's API key permissions and rotate/revoke keys if compromised. If you want the agent to run installs or receive API keys, only proceed if you trust the agent's environment and policies for handling secrets.

Hopkin CLI Skill

When to Use

Use this skill when the user:

• Asks about ad accounts, campaigns, ad sets, ads, or ad performance
• Wants to query Meta Ads, Google Ads, LinkedIn Ads, or Reddit Ads data
• Needs cross-platform advertising reports or comparisons
• References Hopkin, MCP ad servers, or ad platform APIs
• Asks to list, inspect, or analyze advertising entities

Do NOT use this skill for:

• General marketing questions unrelated to ad platform data
• Creating or modifying campaigns (Hopkin is read-only)
• Non-advertising analytics (website analytics, SEO, etc.)

Setup

Installation

Before using Hopkin, ensure it's installed and up to date:

# Install globally from npm
npm install -g @hopkin/cli

# Or if already installed, check if an update is needed (do this if last check was >7 days ago)
npm outdated -g @hopkin/cli && npm install -g @hopkin/cli@latest

Authentication

Check if the user is authenticated before making any data requests:

hopkin auth status --json

If not authenticated, ask the user for their API key:

hopkin auth set-key <API_KEY>

Users can get an API key from https://app.hopkin.ai/settings/api-keys

Discover Available Commands

After authenticating, run hopkin --help to see all platforms, commands, and global flags. This also automatically refreshes the tool cache if it's stale (cache lasts 7 days).

hopkin --help

Then drill down into a specific platform to see its resources and commands:

hopkin meta --help
hopkin google --help
hopkin linkedin --help
hopkin reddit --help

And further into a resource to see its verbs and a specific command to see its flags:

hopkin meta campaigns --help
hopkin meta campaigns list --help

Always use --help to discover what's available before guessing command names — platforms differ in their resource naming and available tools.

How to Use

Query Data

The CLI follows a consistent pattern:

hopkin <platform> <resource> [verb] [flags]

• platform: meta, google, linkedin, reddit
• resource: ad-accounts, campaigns, adsets, ads, etc.
• verb: list, get (defaults to list if omitted)
• flags: --reason (required for most commands), --account, --limit, etc.

Common Commands

# List ad accounts
hopkin meta ad-accounts list --reason "checking accounts" --json
hopkin google accounts list --reason "listing accounts" --json

# List campaigns for an account
hopkin meta campaigns list --account <ACCOUNT_ID> --reason "reviewing campaigns" --json

# Get performance insights
hopkin meta insights --account <ACCOUNT_ID> --reason "performance review" --json

# Ping a platform (health check)
hopkin meta ping --reason "connectivity test" --json

Format Output

Always use --json when processing data programmatically. Other formats are available:

# JSON (best for parsing)
hopkin meta ad-accounts list --reason "..." --json

# CSV (for spreadsheet export)
hopkin meta ad-accounts list --reason "..." --format csv

# TSV
hopkin meta ad-accounts list --reason "..." --format tsv

# Filter specific fields
hopkin meta ad-accounts list --reason "..." --json --fields id,name,currency

# Fetch all pages (for large result sets)
hopkin meta ad-accounts list --reason "..." --json --all

Important Notes

• Read-only: Hopkin cannot modify campaigns, change budgets, or edit ads. It only reads data.
• --reason flag: Most commands require --reason "..." for audit trail purposes. Always include a brief, honest description of why you're querying the data.
• Account IDs: Many commands require --account <ID>. List accounts first to discover available IDs.
• Pagination: Results are paginated by default (25 per page). Use --all to fetch everything, or --limit N to control page size.
• Caching: The MCP servers cache data. Pass --refresh to a command if the user needs real-time data.
• Platform differences: Not all platforms have the same resources. Google has accounts (not ad-accounts), LinkedIn has campaign-groups and creatives (not adsets and ads). Always check hopkin <platform> --help first.

Update Check

Before each session where you use Hopkin, check if the CLI needs updating. Do this if you haven't checked in the last 7 days:

hopkin --version
npm view @hopkin/cli version

If the npm version is newer, update:

npm install -g @hopkin/cli@latest

Troubleshooting

# Check auth status
hopkin auth status --json

# Debug a failing command
hopkin meta ad-accounts list --reason "debug" --json --debug

# Re-authenticate
hopkin auth set-key <NEW_KEY>

# Re-discover available commands (also refreshes tool cache)
hopkin --help