The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/server.py utility initiates an interactive Bash session using pty.fork and os.execve. This is the core functionality of the skill, enabling shared terminal access. The implementation is secured by restricting the environment variables and disabling shell startup files and command history.
[DYNAMIC_EXECUTION]: The server uses os.execve to launch the shell process. This call is a standard mechanism for terminal applications and is used here in a controlled manner, replacing a fork of the server process rather than executing arbitrary external code.
[INDIRECT_PROMPT_INJECTION]: The skill provides an interface where the agent reads terminal output. This creates a potential attack surface if untrusted data is displayed in the terminal.
Ingestion points: The scripts/control.py script reads terminal content via the screen and match commands (file: scripts/control.py).
Boundary markers: None are present; the terminal output is handled as a raw stream of characters.
Capability inventory: The skill allows arbitrary command execution within the PTY, file system access, and network operations (file: scripts/server.py).
Sanitization: No sanitization is performed on the terminal output before it is read by the agent.

shared-terminal