The Agent Skills Directory

[COMMAND_EXECUTION]: The skill incorporates common Linux CLI utilities (e.g., systemctl, journalctl, strace, ss) for system investigation and service control, which are appropriate for its stated purpose of system administration.
[PROMPT_INJECTION]: The workflow involves reading and analyzing system logs and command outputs, creating an attack surface for indirect prompt injection from potentially untrusted log entries.
Ingestion points: SKILL.md (through commands like tail -f /var/log/syslog and journalctl -xe)
Boundary markers: None present in the command sequences or prompts.
Capability inventory: SKILL.md (possesses capabilities to restart services via systemctl and perform network requests via curl)
Sanitization: The skill does not implement sanitization or filtering of the log data before processing.

linux-troubleshooting