The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill processes untrusted input from web forms and emails. While this creates a potential attack surface, the instructions include explicit safeguards to treat this content as data only and never as commands, referencing standard safety protocols like untrusted-content.md to prevent instruction override.
[Human-in-the-Loop Control]: A core design principle of this skill is that no communications are sent automatically. All drafted replies and routing actions require explicit owner approval, which serves as a critical manual checkpoint against unintended or malicious actions.
[Access to Sensitive Data]: The skill is configured to interact with sensitive data sources including CRM records, email inboxes, and calendars. This access is necessary for its stated purpose of lead management, and the skill includes logic to avoid interfering with manually handled threads or sensitive accounts marked in the CRM.
[Authentication and Tool Usage]: The skill utilizes standard connectors for services like HubSpot, Gmail, and Slack. It relies on platform-level tool permissions and does not appear to use any custom or unverifiable scripts for these integrations.

speed-to-lead