The Agent Skills Directory

[Secure Source Integration]: The skill leverages methodology and documentation from well-known technical organizations. All external references point to official documentation repositories, which are handled as trusted informational sources.
[Responsible AI Vetting]: The MVE coaching methodology includes a dedicated section for Responsible AI (RAI) vetting. It instructs practitioners to evaluate experiments for fairness, reliability, privacy, transparency, and accountability before proceeding, which is a significant security and ethical best practice.
[Absence of Executable Code]: The skill consists entirely of reference material and instructions for textual analysis. It does not contain scripts, package dependencies, or shell commands, effectively eliminating common attack vectors like remote code execution or privilege escalation.
[Controlled Data Processing]: While the skill ingests user-provided content such as business outcomes and problem statements, it uses this data solely to generate structured text artifacts (like MVE plans and backlog briefs). It does not pass this data to sensitive APIs or execution environments, which minimizes the risk associated with indirect prompt injection.

experiment-design