The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script deck.py using python3 located at ../heavy-think/deck.py. This is used to retrieve debate formats and rules.
[DYNAMIC_EXECUTION]: The rules and evidence-backed debate formats are dynamically generated at runtime through the execution of the external deck.py script.
[INDIRECT_PROMPT_INJECTION]: As a moderator skill that synthesizes arguments from multiple sub-agents, it inherently possesses an attack surface for indirect prompt injection from agent outputs.
Ingestion points: Arguments and responses generated by 'teammates' (sub-agents) throughout the debate rounds.
Boundary markers: No specific boundary markers or 'ignore' instructions for teammate content are described in this file.
Capability inventory: Subprocess execution of a local Python script and file system access for writing transcript baselines and verdicts.
Sanitization: The skill does not document any sanitization or validation of the sub-agent outputs before they are processed by the main moderator session.

debate