The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires administrative privileges to function, specifically using 'sudo systemctl start postgresql' to initialize its database environment. Executing commands with 'sudo' bypasses standard user restrictions.
[DATA_EXFILTRATION]: The skill includes instructions for high-impact data exposure and harvesting, including:
Extracting password hashes ('hashdump')
Starting keystroke loggers ('keyscan_start')
Taking screen captures of the target system ('screenshot')
Collecting environment variables and credentials ('credential_collector')
Exporting captured data to local files ('creds -o', 'loot -o')
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through the ingestion of external vulnerability scan results. An attacker who can influence the content of a Nessus scan file could embed malicious instructions that the agent might follow when processing the data.
Ingestion points: The command 'db_import /path/to/nessus_scan.nessus' loads untrusted external data into the active session.
Boundary markers: No specific delimiters or safety instructions are provided to the agent to disregard natural language instructions found within the imported data.
Capability inventory: The skill possesses extensive capabilities including remote code execution, network pivoting (autoroute/socks_proxy), and filesystem writes.
Sanitization: There is no evidence of validation or sanitization for the content of the imported scan files.

exploiting-vulnerabilities-with-metasploit-framework