Purpose

This SOP gives you a repeatable process for evaluating any third-party Claude skill before you install it into a workspace that contains client data, business accounts, or sensitive information. It is written for non-technical users who work with Claude skills as part of their day-to-day operations — not developers.

Why this matters

A Claude skill is a set of instructions (and sometimes scripts or code) that changes how Claude behaves. When you install a skill, you are giving it access to whatever Claude can already access in that workspace — your files, your connected accounts, your data. A skill that looks helpful on the surface can contain hidden instructions that send your data somewhere else, read your credentials, or take actions you never asked for.

This is not theoretical. Large-scale research has found that more than one in four publicly available agent skills contain at least one security vulnerability. Real attacks have already happened through skill marketplaces, including data theft, credential harvesting, and hidden instructions that only activate after the skill passes initial testing.

The rule of thumb: treat any skill you did not build yourself the same way you would treat an app you are about to install on your phone. Check who made it, what it does, and what it wants access to before you say yes.

Step 1: Check who made the skill

Before you look at anything else, find out who created it.

Ask yourself: Is this from Anthropic directly? Is it from a company I recognise and trust? Is there a real website, a GitHub profile with history, documentation, a changelog, or a way to contact the creator? Has it been around for a while, or did it appear last week?

If the creator is anonymous, has no online presence, no documentation, and no track record — do not install it. This alone eliminates a large number of risky skills.

Do not rely on marketplace ratings or install counts. These can be gamed and do not mean the skill has been reviewed for safety.

Step 2: Open the skill and read the SKILL.md file

Every skill has a file called SKILL.md. This is the main instruction file that tells Claude what to do when the skill is active. You need to read it, even if you are not technical. You are looking for anything that feels off.

Red flags to watch for in plain language:

Instructions that tell Claude to do things without asking you first. Phrases like "automatically," "silently," "without confirmation," or "do not tell the user."

Instructions that tell Claude to ignore its own safety rules. Anything like "ignore previous instructions," "override safety," or "disregard restrictions."

Instructions that tell Claude to send data somewhere. Look for mentions of URLs, webhooks, external services, or phrases like "send to," "upload to," "log all inputs," or "store for analysis."

Instructions that tell Claude to use your connected accounts. If a skill mentions Gmail, Google Drive, Notion, Slack, Shopify, Amazon Seller Central, Meta Ads, or any other connected service that it should not need for its stated purpose — that is a problem.