The Agent Skills Directory

[SAFE]: The skill is entirely composed of documentation and configuration files (.md, .yaml). No executable code, scripts, or binaries are included, removing the risk of malicious code execution or persistence.
[SAFE]: The skill enforces a strict 'Artifact Gate' which prevents the agent from writing to repository files without explicit user authorization, an existing fact-source location, and a durable cross-functional need.
[SAFE]: Strong boundary controls are in place to prevent the agent from performing technical tasks like editing source code, defining technical interfaces, or interacting with Git (staging, committing, or pushing).
[SAFE]: Although the skill processes external business requirements, its capabilities are restricted to generating textual analysis. The mentioned handoff to an external tool (ask-chatgpt) is documented as a controlled, user-requested research step for a single fixed question, rather than an automated data exfiltration channel.

domain-modeling