storage-san-attack
Fail
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: CRITICALINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and process data from untrusted external sources, creating a potential attack surface for indirect prompt injection.
- Ingestion points: The agent is directed to use tools like
nmap,curl,snmpwalk, andaws s3 lsto gather information from remote storage appliances and cloud buckets. - Boundary markers: There are no explicit instructions for the agent to use delimiters or ignore embedded instructions when processing this external output.
- Capability inventory: The skill has access to the
Bashtool, allowing it to execute shell commands based on its analysis of the ingested data. - Sanitization: The instructions do not specify any validation or sanitization steps for the data retrieved from external systems before it is processed by the agent.
- [EXTERNAL_DOWNLOADS]: The skill's documentation and guides suggest downloading several security tools and utilities from external sources.
- Evidence: The guides mention downloading
minioandmcfromdl.min.io,aws-vaultfrom its GitHub release page, and various pentest tools likelazys3,S3Scanner, andslurpfrom GitHub repositories (e.g.,github.com/sa7mon/S3Scanner). These references are documented neutrally as part of the lab setup and tool inventory. - [COMMAND_EXECUTION]: The skill provides numerous examples of shell commands for the agent to use when interacting with storage systems.
- Evidence: The
payloads.mdandtest-cases.mdfiles contain a wide array of commands for iSCSI login, NFS mounting, SMB relay, and S3 bucket manipulation. These include commands likeiscsiadm,mount,crackmapexec, andaws s3api. These are legitimate tools for the intended purpose of the skill.
Recommendations
- Contains 2 malicious URL(s) - DO NOT USE
Audit Metadata