The Agent Skills Directory

Autonomous Data Analysis (Indirect Prompt Injection Surface): The skill is designed to autonomously analyze site configuration, source code, and Dataverse metadata. While this is essential for its function, it creates a surface for indirect prompt injection where instructions embedded in the analyzed data could attempt to influence the agent's behavior. This is a common consideration for audit tools and is mitigated by the skill's read-only analysis phase and the use of the highly capable Opus model.
Privacy-Preserving Telemetry Design: The skill implements telemetry for tracking audit completion stages but explicitly forbids the transmission of sensitive information such as file paths, table names, or error messages. This use of anonymized identifiers is a positive security practice that protects user and site data.
Authorized Vendor API Integration: To perform its audit, the skill interacts with Microsoft Dataverse APIs to retrieve relationship and metadata information. These operations are conducted using the user's authenticated environment and target official Microsoft endpoints, aligning with expected and documented functionality.
Human-in-the-Loop Remediation: Although the skill can identify security issues, any automated fixes are performed by a specialized sub-agent only after explicit user confirmation. This ensures that the user maintains full control over any changes to the site's security configuration.

audit-permissions