The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill recommends downloading and installing a third-party tool from a non-vendor GitHub repository (https://github.com/mxrch/snmp-shell). This process involves executing git clone followed by pip install -r requirements.txt and running the script, which introduces a supply chain risk to the agent environment.
[DYNAMIC_EXECUTION]: Provides detailed payloads for generating and executing dynamic code on target systems. Specifically, it describes using snmpset to inject and trigger a Python-based reverse shell via the NET-SNMP-EXTEND-MIB MIB extension.
[DATA_EXPOSURE]: Includes instructions for exfiltrating sensitive configuration files from Cisco networking equipment. The methodology leverages snmpset to command the device to upload its running-config to a remote TFTP server, which typically contains plaintext or reversible credentials.
[REMOTE_CODE_EXECUTION]: Instructs the agent to execute a variety of powerful command-line tools such as nmap, snmpwalk, onesixtyone, hydra, and snmpset for interacting with and exploiting target network services.
[INDIRECT_PROMPT_INJECTION]: The skill implements a workflow that requires the agent to parse and interpret untrusted data from external network targets.
Ingestion points: Network responses captured from snmpwalk, snmp-check, and braa execution against target IPs.
Boundary markers: Absent. The instructions do not define delimiters or provide warnings to the agent regarding the potential for malicious data in MIB outputs.
Capability inventory: The environment possesses capabilities including network scanning, arbitrary command execution, and package installation.
Sanitization: None. Data from external targets is processed directly without filtering or verification.

snmp-pentesting