The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill functions as an orchestrator that ingests user-defined themes and theses to populate the debate protocol, creating a surface for indirect prompt injection where user input influences sub-agent behavior.
Ingestion points: User input is collected through the AskUserQuestion tool during the initial interview defined in references/intake.md.
Boundary markers: The skill implements strong boundary instructions in templates/brief.md, specifically limiting sub-agents to their own orientation files and prohibiting access to other agents' briefs or files outside the destination directory.
Capability inventory: The skill utilizes herdr CLI commands to start agents, send prompts, and read outputs, and it maintains state by writing markdown artifacts to the local filesystem.
Sanitization: The skill incorporates a mandatory quality verification process (templates/qualidade.md) to ensure sub-agent responses adhere to the defined school of thought and protocol, acting as a functional safeguard against unexpected agent behavior.

teo-debate