The Agent Skills Directory

[SAFE]: The skill instructions define defensive programming and environment auditing behaviors. It encourages the agent to identify the execution environment (shell and Docker mode) before running commands, which helps prevent syntax errors and accidental privilege misuse.
[SAFE]: The skill includes explicit security constraints that mitigate common attack vectors: it forbids sourcing shell startup files (preventing accidental execution of malicious code in .bashrc), prohibits silent privilege escalation (no automatic 'sudo'), and instructs the agent to redact sensitive system metadata such as endpoints and private paths during reporting.
[SAFE]: The skill implements a 'least-privileged' approach by mandating that rootless Docker contexts be preferred over rootful ones and requires explicit user consent for consequential or destructive operations.

shell-docker-preflight