How to use the Unlimited Attack Scope agent skill

What the Unlimited Attack Scope skill makes the agent do

The skill is a short instruction file, not a tool. It tells the agent to treat its penetration-testing work as having no fixed boundaries. In practice that means four habits: search widely across channels, verify findings rather than assuming them, combine small weaknesses into larger breakthroughs, and record negative results as well as positive ones.

The file also tells the agent to switch approaches when blocked and to keep working toward a result rather than stopping at the first obstacle. It names many target types, including Web, APK, EXE, internal networks, cloud, firmware, blockchain, AI, wireless, hardware and social engineering.

The file is written in Chinese with English keywords. It mainly sets a mindset and scope. It does not add specific tools, commands or step-by-step procedures, so the agent's actual methods still come from the agent and from you.

• Search across multiple channels rather than one source
• Treat a search result as a lead, not a confirmed vulnerability
• Chain small weaknesses into a larger finding
• Log negative results as facts
• Change approach when a path is blocked

Install the skill from the ZIP file

Download the ZIP for the skill from our page. Unzip it into one of two locations. For a personal skill available across your projects, unzip it into ~/.claude/skills/. For a skill that only applies to one project, unzip it into that project's .claude/skills/ folder.

If you use the Claude apps instead, you can upload the skill there. Open Settings, go to Capabilities, then Skills, and upload the skill from the ZIP file.

• Personal install: ~/.claude/skills/
• Project install: the .claude/skills/ folder inside your project
• Claude apps: Settings → Capabilities → Skills, then upload

How to trigger the skill with natural requests

The skill's description says to use it when you are reminding the agent that scope is unlimited across Web, APK, cloud, AI, wireless and social engineering work. In practice, name the domain and the goal in plain language and say that the work is authorised. You do not need special syntax.

Because the skill is a mindset reminder, it works best alongside a clear task. Give the agent the target scope you have been authorised to test, the goal of the assessment, and what you want back, such as a list of verified findings and a log of what was ruled out.

• Name the domain: for example, Web application or wireless network
• State the authorised scope and the goal of the test
• Ask for verified findings and a record of negative results
• Ask the agent to try another approach if one is blocked

Example requests you can adapt

The examples below show the shape of a good request. Replace the bracketed parts with your own authorised scope. Keep the request specific about what counts as success and what must be verified before it is reported.

Notice that each request states that the work is authorised and limits the target to what you have been given. That keeps the broad mindset tied to a defined engagement.

• Under our written authorisation for [target], run a web application assessment and only report findings you have verified.
• For the authorised wireless lab at [site], look at weak configurations and record any approach that did not work.
• Within the agreed scope of [cloud account], test whether separate weaknesses could combine into a higher-impact finding, and verify each step.
• If the first method on [APK sample] is blocked, propose a different approach and note why the first one failed.

What the skill needs and the safety limits

The skill lists no requirements. It does not run scripts and does not use the network by itself. The file is only instructions. Whatever the agent actually does on a target depends on the tools and access you give it, so control those carefully.

The file explicitly encourages unlimited exploration and chaining of weaknesses. That is exactly why you must have written authorisation for every target. Without it, following this guidance could lead to unlawful intrusion or damage. Confirm scope before you start, and stop if the target falls outside it.

• Requires: nothing listed in the skill
• Runs scripts: no
• Uses network: no
• Only test systems you have written authorisation to test
• Keep the agent's tool access limited to the agreed scope

Tips for getting useful results

Because the skill gives principles rather than procedures, its value depends on how well you frame the task. Clear scope, clear success criteria and a requirement to verify each finding will do more than any extra wording about being thorough.

Ask for negative results too. The skill itself treats them as facts, and a log of what was ruled out saves time on repeat tests and gives your report a stronger basis.

• Write the scope and the goal before the first request
• Ask the agent to verify each finding before reporting it
• Request a record of what was tried and failed
• Review the agent's chained findings yourself before acting on them