The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by interpolating untrusted data into prompts for background sub-agents (Skeptic, Pragmatist, and Critic).
Ingestion points: The variables [DECISION QUESTION] and [COMPACT CONTEXT] are directly inserted into the instructions for three parallel agents within the task blocks in SKILL.md.
Boundary markers: The prompt templates lack explicit boundary markers (such as XML tags or dedicated delimiters) or defensive instructions to prevent the interpolated data from overriding the sub-agent's persona or instructions.
Capability inventory: The skill utilizes the platform's background task orchestration to spawn additional agents and consumes their output via read_agent for synthesis.
Sanitization: There are no visible steps to sanitize, escape, or validate the content of the decision question or context before it is passed to the sub-agents.

council