http-host-header-attacks
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The methodology involves the agent interacting with external web servers and interpreting their responses, which presents a surface for indirect prompt injection if the external server returns malicious instructions within HTTP headers or bodies. * Ingestion points: Analysis of HTTP responses and headers (SKILL.md, references/host-header-exploitation.md). * Boundary markers: Not implemented; no specific instructions to isolate or ignore instructions embedded in target data. * Capability inventory: Shell command execution capabilities using tools like
curl,ffuf, andcrlfuzz. * Sanitization: Absent; no procedures provided to validate or escape data retrieved from external targets. - [COMMAND_EXECUTION]: The skill provides functional shell commands for the agent to perform penetration testing tasks, specifically using
curl,ffuf, andcrlfuzzto test for header-based vulnerabilities and enumerate virtual hosts.
Audit Metadata