The Agent Skills Directory

Instructional Content: The file contains purely descriptive and instructional content regarding the project's internal shell execution architecture. It does not include any scripts, executable commands, or remote network operations.
Binary Stream Labeling: The skill mentions specific binary prefixes (\x01\x01\x01 and \x02\x02\x02) used to differentiate between stdout and stderr. These are used as data markers within the system's logging protocol and do not represent obfuscation or hidden malicious instructions.
Command Execution Logic: While the skill discusses shell process management (exec, execStream), it describes how these components function within the 'sandbox-container' project rather than directing the agent to execute arbitrary or unsafe commands on the host system.
Resource Management Guidance: The instructions emphasize security-positive practices such as atomic file operations (using temporary files and moves) and proper resource cleanup (FIFO cleanup in error paths), which are standard for building robust systems.

session-execution