linux-disk-storage

Pass

Audited by Gen Agent Trust Hub on Oct 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONPRIVILEGE_ESCALATIONEXTERNAL_DOWNLOADSPERSISTENCEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides numerous instructions for executing high-risk system commands using sudo, including fdisk, mkfs, truncate, wipefs, and badblocks -w. These tools are necessary for disk management but can lead to irreversible data loss if applied to the wrong device.
  • [PRIVILEGE_ESCALATION]: Nearly all operations described in the skill require sudo privileges, including creating swapfiles, modifying /etc/fstab, and managing network mounts. While consistent with the skill's purpose, it relies on the agent having extensive administrative control over the host.
  • [EXTERNAL_DOWNLOADS]: The skill automates the installation of system utilities such as cifs-utils, smbclient, nfs-common, and autofs using standard package managers (apt and dnf) from official repositories.
  • [PERSISTENCE]: The skill manages system persistence by modifying /etc/fstab to ensure network shares and swapfiles are mounted at boot. It also provides instructions for configuring autofs and systemd mount units.
  • [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection via the processing of untrusted filesystem metadata.
  • Ingestion points: Metadata and filenames ingested through commands like df, du, lsblk, findmnt, and find (documented in SKILL.md and references/storage-reference.md).
  • Boundary markers: Absent; the skill processes standard shell command output without specific delimiters or warnings for the agent regarding embedded content.
  • Capability inventory: The skill possesses extensive capabilities for file system mutation, including rm, truncate, mkfs, and mount (in SKILL.md and scripts/sk-cifs-mount.sh).
  • Sanitization: Absent; the provided scripts and instructions do not include specific sanitization logic for metadata retrieved from the filesystem before it is processed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 6, 2026, 03:13 AM