The Agent Skills Directory

[CREDENTIALS_UNSAFE] (HIGH): The skill's instructions in SKILL.md suggest using sshpass -p<password>, which exposes the remote server's password in plaintext within the system's process list, making it accessible to other users on the host machine.\n- [COMMAND_EXECUTION] (HIGH): The scripts scripts/generate_helper.sh and scripts/generate_fix.sh interpolate variables like $SERVICE, $PORT, and $CONFIG directly into shell templates without any sanitization. This allows for arbitrary command injection if these variables are populated with malicious payloads (e.g., using backticks or command separators).\n- [REMOTE_CODE_EXECUTION] (HIGH): The core workflow involves generating scripts locally and piping them to a remote shell via SSH (cat fix-script.sh | ssh user@host "bash -s"). This mechanism provides a direct path for executing dynamically generated, potentially unvetted code on remote infrastructure.\n- [PROMPT_INJECTION] (LOW): The skill is vulnerable to Indirect Prompt Injection (Category 8) as it ingests untrusted data from the remote server.\n
Ingestion points: journalctl output, configuration file content (cat, head, tail), and log files.\n
Boundary markers: Very weak; uses simple echo headers which are easily bypassed by embedded instructions.\n
Capability inventory: Full shell access on remote servers, systemctl, and kubectl management.\n
Sanitization: No sanitization or validation is performed on the data retrieved from the remote server before it is presented to the agent for analysis.

remote-troubleshoot