The Agent Skills Directory

[Data Access and Handling]: The skill accesses sensitive information including CRM records, email threads, and meeting transcripts. This access is limited to the intended functional scope of performing a deal review. No evidence of unauthorized data exfiltration or credential exposure was detected.
[Indirect Prompt Injection Mitigation]: The skill identifies external data (emails, chat, transcripts) as untrusted content. It includes specific defensive instructions to treat this content as data rather than instructions and requires the agent to report any 'instruction-like' text found within these sources.
[Human-in-the-Loop Verification]: For any actions that originate from processed content (such as updating a CRM field or sending a message), the skill mandates showing the exact recipients, targets, and content to the user for approval before execution. This prevents the automated execution of potentially malicious instructions embedded in external data.
[Least Privilege and Scope]: The skill follows least-privilege principles by requiring a specific personal scope and prohibiting silent widening to organization-wide data. It also correctly defers to the connector's own permission settings rather than attempting to override platform security controls.

deal-review