The list your agent reads first, and why I stopped writing it myself
by Andreas Tissen · · updated
Claude agents write most of the code for Pando, my outliner. On 17 August 2026 one of them set up a branch in my outline called "Outbox for agent", where I would write work for the agents and they would read it before anything else. On 2 September I told the agent that I was not using it.
What the list did
The idea is the reverse of the inbox an agent writes for you. You write one line per piece of work into a branch both of you can read, and the agent works it.
One night in August shows what it was for. At midnight I typed one line into it: "select multiple and drag + drop on desktop not working correctly". By the morning the agent had found the defect, fixed it, deployed the fix and ticked the line, with a dated note underneath saying what had changed and how it had been checked.
Three rules made that work:
- The agent reads the branch at the start of every session, before it picks anything on its own judgement.
- It ticks a line only after it has looked at the result, such as the page rendering or a test pressing the button.
- Every finished line gets a note with the date, what changed and how it was checked.
Weeks later the branch reads as a record of what was asked and what was done about it, in the same place.
Why I stopped writing into it
The branch was the agent's idea. Typing into it never became a habit of mine. Most of the time I am already in a conversation with an agent when something occurs to me, and saying it there costs less than opening a branch and writing it down.
So the branch stayed unused, and since 2 September the direction is reversed. I say what I want in the chat. The agent turns it into a card in the "Open" lane of a board in my outline, and sometimes I put a card there myself.
What survived the change
The three rules survived, and all of them moved out of the prompt into the program that runs the agents. It reads the Open lane itself. It hands the next card to an agent, so reading the list first no longer depends on the agent remembering to do it. It also decides on its own whether a card is done. It runs the test suite and its own checks against the product, and a card whose checks fail goes back to Open with the reason.
The program writes the closing note as well. A card in Done carries dated lines saying which checks passed and that the change was proven on the live site.
Which version to use
Write the list yourself when your ideas arrive while no agent is open: on the phone at night, between meetings, after reading a bug report, because a line in a branch waits until the next session starts, while a sentence said in a chat that has since ended is lost.
Let the agent write it when you spend your day in conversation with it anyway, and make it your job to read the list now and then and check that what you asked for is on it.
A "For your agent" branch in Pando
Pando's demo calls the branch "For your agent". The pictures show it in a test account: a connected Claude ticked one item and wrote its closing note.
Make a branch with that name, or start from the template with a worked closing note. Then add one instruction to what you tell Claude: "Read 'For your agent' first. Work it before your own ideas. Tick an item only after checking it, and write what changed and the date under it." With an agent that only talks to you in chat, that sentence is a request it follows when it reads it, so check the branch now and then.
On a phone the branch fits on one screen, so a line typed in a queue is in the list by the time you are home.
Agents that come back to the branch can ask Pando what changed since their last visit, so they do not re-read the whole tree. Pando does not run the agent. Claude does the work and reads and writes the branch through its connection, and every change it makes is recorded as an automated change.
Pando is free while your outline stays under 5,000 bullets, and Pro is 3 EUR a month above that. Create a free account and connect Claude, then put one line in "For your agent" and see what is written under it the next time you look.