The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous shell commands for running security tools like responder, ntlmrelayx.py, and hashcat, including instructions to use interactive_session for persistent command execution.
[EXTERNAL_DOWNLOADS]: References the official repository for the Responder tool on GitHub (github.com/lgandx/Responder).
[INDIRECT_PROMPT_INJECTION]: The skill sets up an ingestion path for untrusted data captured from the local network, which is then stored in log files and read by the agent, creating a vulnerability surface.
Ingestion points: Captured network traffic and credentials stored in log files such as /usr/share/responder/logs/.
Boundary markers: None identified; there are no instructions to use delimiters or ignore potentially malicious strings in the captured data.
Capability inventory: The agent can execute shell commands and read/write files via interactive_session.
Sanitization: No sanitization or escaping of the captured hostnames, usernames, or other network strings is mentioned.

responder-poison