The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection by design, as it instructs the agent to ingest and analyze untrusted external data from system logs and diagnostics.
Ingestion points: The skill frequently uses journalctl, dmesg, and ausearch to gather system state and triage failures (found in SKILL.md, references/systemd-and-journal.md, and references/security-and-updates.md).
Boundary markers: There are no explicit boundary markers or instructions to ignore embedded commands within the log data being processed.
Capability inventory: The skill has high-privilege capabilities including package management (dnf, rpm), service control (systemctl), firewall modification (firewall-cmd), and SELinux policy management (semanage), often executed with sudo (found throughout the reference files).
Sanitization: No sanitization or filtering of the ingested log content is specified before the agent processes it for decision-making.
[COMMAND_EXECUTION]: The skill makes extensive use of administrative shell commands, including those requiring elevated privileges via sudo for system management tasks.
Evidence: Commands such as sudo dnf remove, sudo leapp upgrade, sudo semanage port, and sudo firewall-cmd --reload are provided as standard operational procedures in references/packages-and-repos.md and references/security-and-updates.md.
Context: These operations are consistent with the skill's primary purpose of RHEL/Fedora administration and are accompanied by warnings to verify transactions and confirm consequential changes before execution.

rhel-fedora