The Agent Skills Directory

[COMMAND_EXECUTION]: The restore_backup function in scripts/db_backup.py is vulnerable to path traversal. The filename parameter is used to construct a file path (self.backup_dir / filename) without validation, which could allow an attacker to read or manipulate files outside the designated backup directory.
[COMMAND_EXECUTION]: In scripts/db_backup.py, the _restore_mongodb method uses shutil.unpack_archive on backup files. This is susceptible to ZipSlip attacks, where a crafted archive containing directory traversal characters could overwrite sensitive system files during extraction. Furthermore, the subsequent shutil.rmtree call on the extraction path could be exploited to delete arbitrary directories if the path was successfully manipulated.
[INDIRECT_PROMPT_INJECTION]: The scripts/db_performance_check.py script retrieves and displays slow query logs from the database. Since these logs include raw query strings that are not sanitized before being presented to the AI agent, they represent an indirect prompt injection vector. An attacker with database access could craft a slow query containing malicious instructions (e.g., in a comment) designed to influence the agent's behavior when it analyzes the performance report.

databases