The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for executing shell commands and interactive programs within tmux sessions using send-keys and run-shell in SKILL.md. This functionality is central to the skill's purpose of terminal management.\n- [INDIRECT_PROMPT_INJECTION]: The skill implements a pattern where the agent reads and processes output from terminal panes, creating a surface for indirect prompt injection if the captured output originates from an untrusted process.\n
Ingestion points: Terminal output captured via tmux capture-pane -p as described in SKILL.md.\n
Boundary markers: The sentinel echo pattern (__DONE__) is used for completion detection but does not distinguish between program output and the sentinel itself.\n
Capability inventory: The agent can execute commands via tmux send-keys and run-shell in SKILL.md.\n
Sanitization: No sanitization or escaping of the captured output is demonstrated before the agent processes it.

tmux-skill