The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands to implement safety hooks and auditing mechanisms. It provides templates for the agent to inspect its own tool inputs using utilities like grep to block high-risk operations such as recursive deletions (rm -rf) or unauthorized git pushes. These commands are defensive in nature and intended to restrict the agent's capabilities to a safe subset.
[SAFE]: The skill's primary objective is to enhance the security posture of an AI agent environment. It advocates for the principle of least privilege, recoverable mistakes via git checkpoints, and transparent auditing. It includes specific guidance on secret management (e.g., removing .env files from git tracking) and emphasizes human-in-the-loop oversight for destructive actions. No evidence of prompt injection, data exfiltration, or obfuscation was detected.

nv-guard