The Agent Skills Directory

[COMMAND_EXECUTION]: The skill constructs shell commands to invoke audience-simulator.py, passing user-supplied "stimulus" and "questions" as arguments. This pattern is susceptible to command injection if the input contains shell-sensitive characters like semicolons, backticks, or pipes.
[DYNAMIC_EXECUTION]: The skill performs runtime assembly of executable commands using python "${CLAUDE_PLUGIN_ROOT}/scripts/audience-simulator.py" with arguments derived from external user input.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data (stimuli and questions) which are then processed by AI personas, potentially influencing the agent's behavior or output structure.
Ingestion points: The stimulus and questions fields provided by the user (as described in SKILL.md).
Boundary markers: None are specified to delimit user content from instructions.
Capability inventory: The skill executes shell commands via python scripts and reads/writes brand profile files.
Sanitization: No sanitization or escaping of the stimulus or questions is described before they are passed to the script or the personas.

focus-group