The Agent Skills Directory

[Unverifiable Dependencies & Remote Code Execution] (CRITICAL): The skill instructs the agent to execute a remote script directly from the internet using a piped shell.
Evidence: The 'Recovery After Complete Failure' section contains `curl -sfL https://get.k3s.io | ... sh -s
server`, which is a classic piped remote execution pattern targeting an unverified (non-trusted) source.
[Privilege Escalation] (HIGH): The skill performs numerous operations requiring root privileges and modifies sensitive system paths.
Evidence: Extensive use of sudo to create scripts in /usr/local/bin/ and /root/.
Evidence: Manual modification of /etc/crontab and /etc/systemd/system/.
[Data Exposure & Exfiltration] (HIGH): The skill accesses and packages sensitive Kubernetes credentials and security tokens.
Evidence: The backup targets /var/lib/rancher/k3s/server/ which contains the cluster 'node-token' and SQLite database, as well as /etc/rancher/k3s/ which contains certificates.
[Persistence Mechanisms] (HIGH): The skill establishes multiple permanent triggers for code execution on the host system.
Evidence: Installation of a systemd service (k3s-pre-shutdown.service) and a root cron job in /etc/crontab.
[Dynamic Execution] (HIGH): The recovery script performs an unsafe system-wide file restoration.
Evidence: The command tar -xzf $BACKUP_FILE -C / in k3s-recovery.sh extracts an archive directly to the root directory, allowing for arbitrary file overwrites if the backup file is malicious or tampered with.

k3s-backup