Building an AI agent that is immediately useful requires providing it with context. This initial knowledge acquisition can be a significant hurdle. For developers and users looking to quickly establish an agent's foundational understanding, the cold-start skill offers a structured approach.\n\ncold-start automates the day-one bootstrapping of an agent's knowledge base. Its purpose is to turn an empty brain into a working system in a single session by carefully sequencing high-value data sources. This means your agent can become functional faster, with relevant information already integrated.\n\n
Understanding How cold-start Functions\n\nThe cold-start skill operates through eight prioritized import phases. These phases are designed to systematically build your agent's understanding from various common data repositories. It begins with local or readily available information and then progresses to more external sources, always with explicit user control.\n\nThe import sequence starts with an existing markdown vault, which often contains personal notes, documentation, or project details. Following this, it integrates data from Google Contacts, providing a network of individuals and their associated details. Next, Google Calendar entries are processed to establish timelines, appointments, and project schedules. Gmail content is then imported to provide communication context and historical discussions.\n\nBeyond these core productivity tools, cold-start also handles conversation exports from messaging platforms, allowing the agent to learn from past interactions. Social media archives contribute to a broader understanding of interests and public engagements. Finally, it incorporates general file collections, ensuring documents and other digital assets are part of the agent's knowledge base. Each of these phases specifically asks for consent before proceeding, putting you in control of what data gets imported.\n\n
Secure Data Integration\n\nData security and user consent are central to cold-start's design. The skill uses a credential gateway for all external service integrations. This means the AI agent itself never directly holds raw OAuth tokens or other sensitive authentication credentials. Your data access is mediated securely.\n\nThis approach enhances security by minimizing direct exposure of your credentials to the agent's core system. For users who prefer not to directly connect services, cold-start offers an alternative path: you can utilize offline exports. Services like Google Takeout provide comprehensive data archives, and other platforms allow for direct archive downloads. These exports can then be imported into the agent, ensuring that data can still be bootstrapped without direct live service connections.\n\n
Practical Application Example\n\nConsider setting up a new AI agent to assist with your professional life. Instead of manually feeding it documents and context, you can simply invoke cold-start with a prompt like 'cold start' or 'bootstrap my data'. The skill will then orchestrate the import process.\n\nFor example, it would first ask to access your markdown notes on ongoing projects. After your consent, it would prompt for Google Contacts to understand your professional network. Next, it would request access to Google Calendar to populate its timeline with your meetings and deadlines. Imagine your agent now having context on your colleagues, your schedule, and project-related communications from Gmail, all within a single session. This initial data infusion provides a robust foundation for the agent to perform queries, retrieve information, and offer informed assistance from day one. Other triggers include 'fill my brain' and 'now what?'.\n\n
Scope and Limitations\n\nIt is important to understand what cold-start is designed for and what it is not. This skill is a one-time setup orchestrator. Its primary function is to establish a foundational knowledge base efficiently. It is not intended for continuous monitoring of data sources or as a replacement for live synchronization services.\n\nOnce the initial bootstrapping is complete, the agent's brain will be populated with the imported data. Subsequent updates or ongoing data integration would typically be handled by other skills or manual inputs, not by cold-start. Its utility lies specifically in taking an empty agent and giving it an immediate, comprehensive context from your existing digital footprint. The tools it uses to achieve this include search, query, get_page, put_page, add_link, add_timeline_entry, and sync_brain.\n\n
FAQ\n\nQ: How long does the cold-start process take?\nA: The skill is designed to complete the bootstrapping of a knowledge base in a single session, making your agent functional quickly.\n\nQ: What if I don't want to connect my Google accounts directly?\nA: You can use offline exports like Google Takeout or other archive downloads and import them into the agent as an alternative to direct service connections.\n\nQ: Is cold-start meant for ongoing data synchronization?\nA: No, it is a one-time setup orchestrator for initial knowledge base population, not for continuous monitoring or live syncing of data sources.\n\ncold-start provides a direct method for quickly equipping your AI agent with a functional knowledge base right from the start. Its focus on controlled imports and secure integration makes it a practical tool for initial agent deployment.
