skills/iuliandita/skills/rhel-fedora/Gen Agent Trust Hub

rhel-fedora

Pass

Audited by Gen Agent Trust Hub on May 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides extensive access to system administration commands required for its primary purpose of managing RHEL-family distributions.
  • Evidence: Frequent use of administrative tools including dnf, rpm, systemctl, grubby, dracut, semanage, and firewall-cmd throughout all reference files.
  • Context: These commands are used as intended for package management, service troubleshooting, and boot configuration.
  • [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection due to its requirement to read and process potentially untrusted external system data.
  • Ingestion points: The agent is instructed to read output from journalctl, ausearch, dmesg, rpm -qi, and dnf info, which may contain attacker-controlled content like malicious package descriptions or log entries.
  • Boundary markers: There are no explicit instructions or delimiters provided to the agent to distinguish between its instructions and the data being processed from system outputs.
  • Capability inventory: The skill possesses high-privilege capabilities including package installation (dnf), bootloader modification (grubby), and security policy changes (semanage, setsebool).
  • Sanitization: No sanitization or validation mechanisms are described for the data ingested from system commands before it is analyzed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
May 6, 2026, 05:53 AM