2.6 Library in Depth: Team Space, Bulk Work, and Knowledge Sources
Lesson 1.8.1 solved "getting the file in". But once Library holds a few thousand files shared by a dozen people, new questions appear:
- Does this go in My Space or Team Space?
- How do hundreds of files come in, and how should the folders be organised?
- Can the agent actually find the file I mean?
- How do I stop it overwriting an original?
This lesson takes them one at a time.
My Space Versus Team Space: Where the Access Boundary Sits
This is the first thing to get straight.
| Dimension | My Space | Team Space |
|---|---|---|
| Ownership | Individual | Team |
| Who can see it | You | Team members with permission |
| Good for | Personal work files, drafts, temporary material | Shared standards, templates, knowledge drafts |
| Agent access | Available once you authorize | Depends on your permissions inside the team space |
The rule: does anyone else need this file? If yes, Team Space. If no, My Space. Both mistakes cost you — personal drafts in Team Space clutter the shared tree, and team standards in My Space are invisible to colleagues.
Bulk Upload and Folder Conventions
With a handful of files, drag and drop is fine. With many, the folder structure decides whether the whole thing is usable.
How to upload: Library accepts batches of files, and whole folders in one go. When migrating existing material, uploading a folder preserves the original hierarchy and saves a great deal of manual tidying.
Folder conventions: the agent navigates by path and file name. A few practices that pay off:
| Practice | Effect |
|---|---|
| Divide the top level by business domain, keep it within three levels | Avoids five-level nesting that neither people nor the agent can navigate |
| Put a date or version in the file name | Sorts by time and separates old from new |
| Use one naming format per document type | The agent finds the whole set in one pass |
| Never use "New folder" or "final_v2" | Those names help no search, human or machine |
A workable starting point: business domain / document type / year, for example Legal/Contract Templates/2026.
See Upload and Organize Resources for uploading, and Files and Folders for creating folders and moving files.
Search and Filters: Making Files Findable
Whether a file can be found depends on two things: whether your naming is searchable, and whether you gave a clear enough clue.
Enter keywords in Library and narrow results by file type. To filter by tags or metadata, first set the relevant fields and values on the files, then select the matching conditions.
When you ask the agent to find a file inside a task, there are three strengths of clue:
| Clue strength | How you write it | When it fits |
|---|---|---|
| Weak | "take a look at my contracts" | Exploratory work; hit rate is unpredictable |
| Medium | "find contracts with 'supplier' and '2026' in the file name" | When you are unsure of the exact file name |
| Strong | "read Library/Legal/Contract Templates/2026/Standard_Terms.pdf" | Known path; the most reliable |
When you know the file, provide its name and full folder path. Check the version after locating it, then ask the agent to read it.
See Search and Preview for searching the interface and Tags and Metadata for field settings. To locate and read a file in a conversation, use the Library file exercise.
Using Library as a Knowledge Source for Skills
This is where Library stops being a filing cabinet and becomes a capability base.
A Skill can carry reference material and template files. When that material comes from Library, the Skill is no longer just a set of instructions — it is a procedure running against your real assets.
For example: a contract-review Skill whose SKILL.md describes the comparison logic, while the baseline it compares against — your company's standard terms — is a file in Library. When the standard terms are updated, you replace the Library file and the Skill itself never changes.
Output a difference report sorted by risk level.
The value of that structure is separation of concerns: the method lives in the Skill (stable long term), the material lives in Library (updated often). Each is maintained independently.
Save Results as New Files
State the input files and output folder in the task, and include a date or version in the result names. For example:
Troubleshooting
| Symptom | Cause | What to do |
|---|---|---|
| The agent read an older version | Several files share a name | Add version numbers to names, or write the full path in the task |
| "No permission to open" | A link does not grant access | Check you are signed in to the right account, and ask the owner to open access |
| The file is found but its content cannot be read | Unsupported format, or the file is a scan | Convert it to a parseable format, or supply a text version |
| Task output never reaches Library | Task outputs and Library are two separate doors | Download the deliverable from 【Task Outputs】, then follow Upload and Organize to upload it to Library |
Summary
The advanced use of Library is turning it from "where you keep files" into "the knowledge base both your team and the agent can rely on". Choose the space, organize folders, provide file names and paths, and save results to the specified destination.
That completes the material layer. The next lesson takes Base further — especially how data should be designed to hold up over the long run.
2.5 Designing a Work: Assemble Skills Into One Business Chain
You installed a few Skills and they work well. But every task still starts with 'which ones do I need this time?'
2.7 Base in Depth: Views, Fields, and Sharing Permissions
Lesson 1.8.2 covered what Base is. This one covers using it well — view choice, field design, and permission boundaries.