SSH Penetration Testing
Fail
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: HIGHPERSISTENCECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PERSISTENCE]: The skill provides explicit instructions to append a public key to a target's
~/.ssh/authorized_keysfile to maintain unauthorized access across sessions. - Evidence:
echo "ssh-rsa AAAAB3..." >> ~/.ssh/authorized_keysin Phase 8. - [DATA_EXFILTRATION]: The skill includes procedures to locate and access sensitive data, including private SSH keys and user command history that may contain credentials.
- Evidence:
find / -name "id_rsa"andcat ~/.bash_history | grep -i passin Phase 8. - [COMMAND_EXECUTION]: The workflow relies on executing a suite of offensive security tools including Nmap, Hydra, Medusa, and the Metasploit Framework to perform network attacks.
- Evidence: Shell command blocks in Phases 1, 2, 4, 6, and 10.
- [INDIRECT_PROMPT_INJECTION]: The skill's workflow ingests untrusted external data, such as target hostnames and credential wordlists, and interpolates them into shell commands and Python scripts.
- Ingestion points: Network targets, username strings, and wordlist file paths (SKILL.md).
- Boundary markers: No delimiters or "ignore instructions" markers are defined for processed data.
- Capability inventory: Direct shell execution of CLI tools and script-based command execution via
paramiko(SKILL.md). - Sanitization: There are no mechanisms described for validating or escaping external inputs before they are used in commands.
Recommendations
- AI detected serious security threats
Audit Metadata