pentest
Fail
Audited by Gen Agent Trust Hub on Feb 20, 2026
Risk Level: HIGHCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- COMMAND_EXECUTION (HIGH): The
reference/FINAL_REPORT.mdfile defines apandoccommand used for report generation. This command interpolates the{Target}variable directly into a bash execution string. If an attacker can influence the target name, they could achieve arbitrary command execution via shell metacharacters. - COMMAND_EXECUTION (MEDIUM): The skill contains a library of penetration testing commands (nmap, ettercap, responder).
SKILL.mdandquickstart.mdfiles instruct the agent to execute these commands, including high-privilege operations like modifying network settings and performing traffic interception. - EXTERNAL_DOWNLOADS (MEDIUM): The skill references external tools and domains, including a C2 example (attacker.com) and external APIs for IP reputation checks.
- PROMPT_INJECTION (LOW): The skill is susceptible to Indirect Prompt Injection (Category 8). Evidence Chain: 1. Ingestion points: Reconnaissance findings in Phase 3. 2. Boundary markers: None. 3. Capability inventory: Subprocess execution and shell command execution. 4. Sanitization: None.
Recommendations
- AI detected serious security threats
Audit Metadata