Let the Agent read and write your material

Chat has a hidden cost: shuttling. Copy material in, copy conclusions out. Once is fine. Every day is a waste.

The skill wires the Agent to your library

With the LLM-notebook skill installed, the Agent can:

• read: search your material by keyword and get the body text plus its source;
• write: turn the result into a note and file it back into your workspace;
• archive: collect the current session itself into the library.

You stop hand-feeding context, because all your material already lives in one place.

A real sequence

Ask it to draft an industry overview from the three research reports you saved last week:

1. Search material → three reports matched.
2. Read the bodies → market size, growth rate, competitive landscape located.
3. Draft the piece → citation markers kept.
4. Write it back → you edit it in the workspace.

You say one sentence, then you review.

Draw the boundaries

Read-and-write access means the permissions have to be obvious:

• It searches what you authorised, not your whole machine.
• Writing back creates a new note; it does not overwrite your work.
• Every citation in the draft can be clicked and checked.

The Agent is not valuable because it thinks for you. It is valuable because it carries and files for you.