The Agent Skills Directory

[COMMAND_EXECUTION]: Instructions for establishing unauthorized interactive access channels across multiple platforms.
Provides one-liners for reverse shells using Bash, Python, and PowerShell in SKILL.md.
Includes configurations for bind shells on both Linux and Windows systems.
[DATA_EXFILTRATION]: Explicit patterns for moving sensitive system data to external listener addresses.
Demonstrates exfiltrating system files such as /etc/passwd via netcat.
Provides commands for exfiltrating database dumps using mysqldump piped directly to network connections.
[REMOTE_CODE_EXECUTION]: Multi-stage payload delivery and execution workflows.
Describes a process in Pattern 2 of SKILL.md where the agent is instructed to download a script from a remote host, modify its permissions, and execute it locally.
[COMMAND_EXECUTION]: Detailed instructions for establishing persistence mechanisms to maintain access across reboots.
Includes a template for creating a systemd service that executes a netcat backdoor on startup.
Provides crontab entries using the @reboot directive to trigger reverse shells.
Documents the creation of Windows Scheduled Tasks (schtasks) designed to run shells with System-level privileges.
[PROMPT_INJECTION]: High vulnerability surface for indirect prompt injection through untrusted input interpolation.
Ingestion points: Target IP addresses, ports, and file paths are ingested as parameters in SKILL.md.
Boundary markers: Absent; no delimiters or instructions are provided to the agent to ignore embedded instructions in the metadata or data it processes.
Capability inventory: The skill leverages subprocess execution (nc, bash, python, powershell), file system writes (systemd/cron), and broad network operations.
Sanitization: Absent; there is no evidence of validation or escaping for parameters before they are interpolated into shell commands.

network-netcat