emergency-response
Warn
Audited by Gen Agent Trust Hub on Jun 16, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill makes extensive use of system utilities (netstat, ps, wmic, tasklist) and shell commands (bash, cmd, powershell) to analyze system state and detect anomalies.
- [DATA_EXFILTRATION]: Instructs the agent to access and read highly sensitive system files and data structures, which is an exposure risk. Examples include: Linux shadow files (/etc/shadow), sudoers configuration (/etc/sudoers), SSH authorized keys, and the Windows SAM registry hive. While intended for auditing, these operations handle raw credentials and sensitive access controls.
- [EXTERNAL_DOWNLOADS]: Contains instructions to install security tools (chkrootkit, rkhunter) using standard system package managers (apt, yum). These are well-known security utilities from trusted system repositories.
- [PROMPT_INJECTION]: Subject to indirect prompt injection vulnerabilities as it directs the agent to ingest and analyze untrusted data sources that could be manipulated by an attacker. This includes: System logs (/var/log/auth.log, Security.evtx), user command history (.bash_history), and arbitrary files in web directories. Evidence: Ingestion points include log files and command output; Capability inventory includes shell execution on dual platforms; Boundary markers and sanitization methods are absent in the provided instructions.
Audit Metadata