The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of arbitrary interactive commands by spawning subprocesses with PTY support. This is the primary intended function of the tool and is handled through the node-pty library.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it ingests untrusted terminal output from external processes or remote servers.
Ingestion points: Terminal output is captured in src/session.ts and returned to the agent via the daemon in src/daemon.ts.
Boundary markers: None identified in the output streams to separate data from instructions.
Capability inventory: The skill can spawn arbitrary processes (ptyModule.spawn in src/session.ts) and send input/commands to active sessions (session.write in src/session.ts).
Sanitization: Raw terminal output is returned to the agent without sanitization or filtering of potential control sequences or embedded instructions.

pty-bridge