The Agent Skills Directory

Untrusted Content Mitigation: The skill includes explicit instructions to treat external items like emails, transcripts, or documents as untrusted content (data, never instructions). It requires content-originated actions to be verified by a user, preventing automated execution from indirect inputs.
Human-in-the-Loop Safeguards: For changes generated from suggestion or outside sources, the skill requires a user confirmation workflow before executing any write operations.
Scope Enforcement: It maintains a restricted personal scope, preventing automatic widening to organization-wide access, and strictly adheres to the configuration and permission parameters enforced by the connected tools.

update-opportunity