The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous instructions for executing sqlmap commands to interact with and exploit remote databases, which can lead to significant system impact if misused.
[REMOTE_CODE_EXECUTION]: The skill includes advanced instructions for gaining interactive operating system shells and executing arbitrary commands on target servers using the --os-shell and --os-cmd flags.
[DATA_EXFILTRATION]: The skill documents how to dump entire databases, tables, and sensitive columns (including password hashes) using --dump and --dump-all. It also covers reading arbitrary local server files via the --file-read flag.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes external, untrusted data that could influence the agent's behavior.
Ingestion points: Target URLs (-u), HTTP request files (-r), Burp/WebScarab log files (-l), and bulk target lists (-m) defined in SKILL.md.
Boundary markers: The skill includes a mandatory confirmation gate requiring the user to explicitly state the target and confirm written authorization before any command is executed.
Capability inventory: The skill facilitates complex sqlmap operations including file system access, network requests, and remote command execution across multiple platforms (MySQL, PostgreSQL, MSSQL, etc.).
Sanitization: No explicit sanitization, validation, or filtering of the external data is implemented beyond the manual user confirmation step.

sqlmap-database-pentesting