hardening
Pass
Audited by Gen Agent Trust Hub on Nov 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The file
detection-signals.mdcontains an extensive collection of shell commands used to audit system configuration, identity management, network settings, and data protection. These include commands to check effective SSH configurations (sshd -T), firewall rules (nft list ruleset), and kernel parameters (sysctl). These commands are specifically designated as read-only and are intended to collect evidence without modifying the system. - [INDIRECT_PROMPT_INJECTION]: The skill's 'Review' mode is designed to ingest and analyze untrusted data, such as system configurations, IaC files, and command outputs. While this creates a potential surface for indirect prompt injection (e.g., instructions hidden in config files), the skill includes mandatory evidence chain requirements and explicit instructions to redact secrets and evaluate data in a read-only context, which mitigates the risk.
- [DATA_EXPOSURE]: The skill provides instructions to inspect sensitive files like
/etc/shadow,/etc/sudoers, and SSH keys to identify security gaps. However, it mandates a strict 'Redact first' rule (R1 inSKILL.md), instructing the agent to never copy actual secrets into its output and only record the location and nature of the exposure. - [SAFE]: The skill references multiple well-known and industry-standard security tools, such as Lynis, OpenSCAP, Prowler, ScoutSuite, and Trivy. These references are appropriate for the skill's primary function as a security hardening assistant and target official or established sources.
Audit Metadata