The Agent Skills Directory

[SAFE]: The skill implements a robust multi-agent architecture where sub-agents (test-design, metrics, sample-size, guardrail) operate within defined boundaries to produce structured experiment designs.
[SAFE]: A dedicated 'critic-agent' serves as a quality gate, enforcing specific constraints such as numeric success thresholds and guardrail definitions before finalizing artifacts.
[SAFE]: Data access is confined to local context files produced by the same author's marketing skill ecosystem, ensuring a consistent and controlled environment for the agent's operations.
[SAFE]: Tool usage is consistent with the skill's purpose, using search and fetch tools for benchmark research and standard file system tools for context gathering, with no evidence of unauthorized network activity.

experiment