The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is primarily designed to facilitate the execution of shell commands and interactive keystrokes within tmux panes using the tmux send-keys command.
Evidence: Examples provided in SKILL.md and tmux/SKILL.md demonstrate sending arbitrary shell commands and Python code to managed sessions.
[PROMPT_INJECTION]: An indirect prompt injection surface is present as the skill captures and processes output from tmux panes without sanitization, which could lead the agent to interpret captured text as instructions.
Ingestion points: Terminal output is retrieved using tmux capture-pane in SKILL.md and within the scripts/wait-for-text.sh script.
Boundary markers: The skill does not implement delimiters or provide explicit instructions to the agent to disregard commands embedded within the captured terminal output.
Capability inventory: The skill has the capability to both read from (capture-pane) and write to (send-keys) interactive shell sessions.
Sanitization: No sanitization, filtering, or escaping is applied to the text captured from tmux panes before it is incorporated into the agent's context.

tmux