How to use the Database Security Assessment agent skill

What the database-security skill makes the agent do

An agent skill is a folder of instructions that an agent such as Claude loads when a request matches the skill's description. This skill covers authorized database security assessment for PostgreSQL, MySQL, MSSQL, MongoDB and Redis. Once it is loaded, the agent works through a checklist: network exposure and TLS, account roles and grantees, access control on sensitive tables, dangerous configuration, audit logging, and backup and snapshot permissions.

The skill tells the agent to set a clear scope first, covering the instances, the account permissions, and whether writes or deletes are allowed. It also instructs the agent to avoid destructive statements on production databases unless they are explicitly permitted. The instructions are written mostly in Chinese with some English headings.

• Checks exposure: listening addresses, security groups, and whether the database is bound to 0.0.0.0
• Checks authentication: default or test accounts and weak or shared logins
• Checks authorization: over-broad permissions and role separation
• Checks dangerous features such as xp_cmdshell, COPY PROGRAM, UDF, file_priv and load_file
• Checks audit logging and backup access control

How to install the database-security skill

Download the ZIP for skills/database-security from this page and unzip it. The skill folder contains SKILL.md and references/db-misconfig-checklist.md.

For a personal install, unzip it into ~/.claude/skills/. For a single project, unzip it into the project's .claude/skills/ folder. Alternatively, upload the skill in the Claude apps under Settings → Capabilities → Skills.

How to trigger the skill with natural requests

You do not need a special command. The agent loads the skill when your request matches its description, which covers authorized database security assessment and misconfiguration review. Name the database and the goal, and state that you have authorization and a defined scope.

Good requests include the database type, the target, and the kind of review you want. Stating the scope in the request helps the agent follow the skill's first step.

• Review our PostgreSQL instance at the staging address for network exposure and over-broad roles. We have written authorization and the scope is read-only.
• Run a misconfiguration review of our MySQL server against the database checklist and list which items fail.
• Check whether our Redis instance is reachable without authentication. This is an authorized audit of our own servers.
• Prepare hardening evidence for our MongoDB cluster for an upcoming compliance check.

What the skill needs before you start

The skill assumes you already have authorized access and a written scope for the target databases. It expects you to have database client command-line tools for connecting and enumerating. Some optional tools are referenced: sqlmap, nuclei, and access to a cloud RDS console for configuration audit. The skill also refers to related skills by path (pentest-tools, cloud-k8s, precedent-pentest, attack-chain), which are not included in these files, so those references may not resolve on your machine.

• Written authorization and a defined scope for each target
• Official database CLI tools for connecting and enumerating
• Optional: sqlmap, nuclei, cloud RDS console access
• Network access to the target, and the skill's related skills if you want the full chain

Safety notes: scripts and network use

This skill does not run scripts. It does use the network, because the assessment connects to the target databases and may use the optional tools above. Only point it at systems you are authorized to test.

The skill is framed around exploitation and lateral movement, so it can cause legal harm if used without permission. Production systems are at risk from stray queries or injection testing, which can corrupt or delete data. Keep the scope read-only unless you have explicitly allowed writes.

• Do not run it without written permission to test the target
• Avoid destructive statements on production databases
• Separate configuration findings from exploitable chains in your report, as the skill's self-check asks

Tips for a useful assessment

Give the agent the checklist's structure in your request so the output maps to your review. The reference checklist is short, so you can also use it directly as a manual audit list. Ask the agent to separate configuration problems from exploitable paths, which the skill's self-check requires.

• Start with exposure and authentication before permissions
• Record the scope in writing before the first connection
• Ask for a checklist-style result so findings are easy to track