The Agent Skills Directory

[SAFE]: The skill implements a comprehensive risk management strategy for AI-driven actions.
[SAFE]: The approval-framework.md defines a structured process for action authorization, classifying tasks from Low to Critical risk and requiring explicit human-in-the-loop confirmation for high-stakes operations like ad spend or bulk data modification.
[SAFE]: The agency-operations-guide.md details a secure credential isolation model where API keys and tokens are stored in per-brand files outside the plugin directory, preventing cross-client data exposure.
[SAFE]: Includes an automated Quality Assurance system documented in eval-framework-guide.md that uses deterministic checks to detect hallucinations and verify factual claims against evidence libraries before publication.
[SAFE]: All external service references and URLs throughout the 33 files point to legitimate and well-known platforms (e.g., Google, Meta, LinkedIn, Schema.org).

context-engine