The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides detailed instructions and examples for executing high-risk system commands throughout its documentation (SKILL.md, references/ceph.md, references/san-nas-iscsi.md, references/snapshots.md, references/zfs-raid.md). These include loading kernel modules (modprobe), formatting disks (mkfs), managing storage pools (zpool, zfs), and configuring storage targets (targetcli, iscsiadm), which require administrative privileges and can impact system stability.
[EXTERNAL_DOWNLOADS]: Numerous reference files (e.g., architecture-patterns.md, code-organization.md) include Python implementations that use 'aiohttp' to fetch content from arbitrary URLs. This pattern is repeated hundreds of times across the filler content, establishing a capability for the agent to perform network requests to external sources.
[PROMPT_INJECTION]: The 'Response Format' section in SKILL.md explicitly instructs the agent to provide 'No preamble. No postamble. No explanations.', which is a behavioral override that limits the agent's transparency and communication with the user.
[OBFUSCATION]: Eight of the reference files contain exactly 150 sections of nearly identical boilerplate text. This redundant content inflates the skill's footprint by thousands of lines, a technique that can be used to hide malicious instructions or evade thorough security reviews by overwhelming the auditor with noise.

storage-infrastructure