The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute local commands using tools like tmux and the subprocess module to drive interactive terminal applications.
[DYNAMIC_EXECUTION]: The instructions provide templates for generating and executing temporary Python and Bash scripts to automate terminal interactions based on runtime conditions.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data by capturing terminal output (using tmux capture-pane or reading from a PTY master file descriptor). If the application under test prints attacker-controlled content, it could influence the agent's subsequent logic.
Ingestion points: Terminal output captured in SKILL.md via tmux capture-pane and PTY os.read in Python snippets.
Boundary markers: The templates suggest waiting for specific strings (e.g., <ready text>) but do not implement robust delimiters to distinguish data from instructions.
Capability inventory: The skill possesses capabilities to execute commands (subprocess.Popen), send keys to terminals (tmux send-keys), and terminate processes.
Sanitization: Python snippets use errors="replace" during decoding, which prevents character encoding errors but does not sanitize the content for control sequences or instructions.

control-cli