Task management
Your agent can create and track to-do lists directly in its workspace. Ask it to “make a list of action items from this research” or “track these competitor features I need to investigate,” and it will create structured task lists you can reference in future sessions. Tasks persist in the agent’s workspace, so you can pick up where you left off. This is useful for multi-session research projects where you need to track what’s been covered and what’s still outstanding.Skill invocation
Your agent can run any skill that exists in its workspace. Skills are reusable custom tools — things like report formatters, analysis templates, or specialized research workflows — that your agent calls by name during conversations. When a skill is invoked, the agent reads the skill’s instructions and executes them within the current session context. This lets you standardize repeatable tasks without re-explaining your requirements each time.File operations
Your agent can read, write, and manage files within its workspace. This includes:- Creating files to store research outputs, reports, or data exports
- Reading files from previous sessions to maintain context
- Updating files as new information comes in
Command execution
Your agent can run data processing and analysis commands within its workspace. This is useful for transforming data, running calculations, or performing operations that go beyond what the AI can do in its reasoning alone. Commands execute in a sandboxed environment within the agent’s workspace, so they can work with files the agent has created or received from data sources.Database operations
Your agent can create, query, and manage SQLite databases within its workspace. This is valuable when you need to store structured research data, run queries across collected findings, or maintain datasets that grow over time.How built-in tools differ from data sources
Built-in tools and data source tools serve different purposes.| Built-in tools | Data source tools | |
|---|---|---|
| Availability | Always on, every session | Controlled by tool preferences |
| Scope | Operate within the agent’s workspace | Reach out to external services and APIs |
| Configuration | No setup required | Some require OAuth or enablement |
| Examples | File I/O, tasks, commands | Reddit search, web scraping, Gmail |

