performing-external-network-penetration-test

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt embeds literal credentials (e.g., 'Spring2025!' in the sprayhound command and 'public' as an SNMP community string) and instructs using them verbatim in attack commands, which requires the model to output secret values directly.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The document contains explicit offensive and post-exploitation instructions (reverse shells/metasploit usage, Meterpreter commands like hashdump/keyscan_start/screenshot, webshell upload techniques, JNDI callback to attacker.com, persistence and pivoting steps, and instructions for credential harvesting/cracking and data exfiltration) which are high-risk backdoor and data-theft patterns that can be abused if used without strict authorization.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.72). The skill’s workflow includes runtime ingestion of outsider-authored free text via “Phase 2/3/4” OSINT/scanning outputs (e.g., curl fetching https://crt.sh/... JSON, shodan search, theHarvester, gvm-cli/nessuscli results, and tool outputs like nikto/gobuster), which are external web/tool responses that can contain arbitrary text and are fed into the agent’s LLM context as results/evidence.