The Agent Skills Directory

[SAFE]: The skill establishes strong data boundaries by explicitly instructing the agent to avoid reading patient data and to focus only on repository configuration and policies during its analysis.
[SAFE]: Clear authentication protocols are defined, prohibiting the use of production secrets in tests and preventing the exposure of tokens or session material within the workflow.
[SAFE]: External network operations via the WebFetch tool are strictly constrained to authoritative Abridge support and product documentation domains.
[SAFE]: Writing and editing operations are subject to strict approval boundaries, requiring manual environment verification and clinical owner confirmation before any artifacts are modified.

abridge-core-workflow-a