The Agent Skills Directory

[DATA_EXFILTRATION]: The skill provides specific methodologies and commands for extracting sensitive information from remote targets. This includes instructions for dumping entire databases, specific tables, and columns (e.g., --dump, --dump-all). Crucially, it documents how to read sensitive files from the target host's filesystem, specifically referencing sqlmap --file-read="/etc/passwd".
[REMOTE_CODE_EXECUTION]: Explicit instructions are provided for gaining unauthorized command execution on target systems. This includes using the --os-shell flag to obtain an interactive operating system shell and the --os-cmd flag to execute individual commands such as whoami on the remote server.
[COMMAND_EXECUTION]: The skill describes methods to modify a remote server's filesystem by uploading arbitrary files. Documentation for the --file-write and --file-dest flags demonstrates how to place files like web shells (e.g., shell.php) on the target server.
[INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface where it processes data from untrusted external targets while possessing high-privilege capabilities.
- Ingestion points: Untrusted target URLs and web application responses are processed by SQLMap based on parameters defined in SKILL.md.
- Boundary markers: Absent. No delimiters or instructions are provided to the agent to ignore potentially malicious content embedded in the target's HTTP responses.
- Capability inventory: Includes extensive file system access (read/write), database exfiltration, and operating system command execution as documented in SKILL.md.
- Sanitization: Absent. There are no instructions for the agent to sanitize or validate the target URLs or the data returned from them before proceeding with exploitation steps.
[EXTERNAL_DOWNLOADS]: Mentions downloading the SQLMap tool from its official repository on GitHub.

SQLMap Database Penetration Testing