The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for executing high-risk SQLMap commands, including the use of --os-shell to obtain an interactive operating system shell and --os-cmd to execute arbitrary commands on the target server.
[DATA_EXFILTRATION]: The skill is designed to automate the extraction of sensitive database content, including full table dumps via --dump, specific credential extraction for username and password columns, and password hash collection using the --passwords flag.
[REMOTE_CODE_EXECUTION]: The inclusion of --file-write and --file-dest flags enables the agent to upload files, such as web shells, to the remote target, facilitating persistent access and remote code execution.
[CREDENTIALS_UNSAFE]: The skill demonstrates methods for accessing sensitive system files, specifically citing the retrieval of /etc/passwd via the --file-read parameter.
[PROMPT_INJECTION]: The skill presents a significant attack surface for indirect prompt injection by processing untrusted external data.
Ingestion points: Target URLs, Burp Suite request files (-r), log files (-l), and bulk URL lists (-m) provided by users.
Boundary markers: Absent; there are no instructions to use delimiters or warnings to ignore embedded content in the processed data.
Capability inventory: Shell execution via sqlmap subprocess calls.
Sanitization: Absent; the skill does not instruct the agent to validate or sanitize external inputs before interpolating them into shell commands, which could allow a malicious user to trigger command injection on the agent's host environment.

SQLMap Database Penetration Testing