The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses tmux send-keys to programmatically inject and execute commands in terminal panes, including directory changes and process execution.
[COMMAND_EXECUTION]: It performs process management by identifying PIDs through pgrep and tmux display-message, then terminating them using kill and kill -9.
[REMOTE_CODE_EXECUTION]: The skill dynamically creates a shell script in /tmp using mktemp and executes it in a new tmux window. This pattern of generating and then executing a script at runtime is a significant security vector.
[PROMPT_INJECTION]: It implements a handoff mechanism where instructions are written to .claude/self-reborn/handoff.md. Upon restart, the agent is directed to read this file and follow the 'Next Steps' defined within it, which could be exploited to persist malicious instructions across session boundaries.

reload