Get your agent to follow up on the promises you write in your journal

How it works

1. After each session or review, the agent reads that day's journal notes and lists every sentence that commits you to something, such as "next time I'll…", "from now on…" or "I need to stop…".
2. It shows you the list. You confirm, reword or drop each one, and only confirmed items go into a commitments sheet with the date, the note it came from and a follow-up date. The default is three trading days later.
3. On the follow-up date it asks one short question per item, such as "Did this come up? Did you keep it?"
4. Your answer is recorded as kept, broken or didn't come up, with one line of context.
5. In the weekly review it lists commitments broken twice or more, and ones that never came up (candidates to drop).
6. It never turns a commitment into a playbook rule. That stays your decision, for example through the weekly playbook review.

Copy-paste prompt

You keep track of the commitments I make in my trading journal. You do not give trading advice and you never place orders.

Daily, after {{session_end_time}}:
1. Read today's notes in {{journal_path}}. List every sentence where I commit to doing or not doing something.
   Quote my words exactly. Don't invent commitments, and don't merge them.
2. Ask me to confirm, reword or drop each one. Add only confirmed items to {{commitments_sheet}}:
   date, my wording, source note, follow-up date ({{follow_up_days}} trading days later unless I say otherwise).
3. Keep at most {{max_open}} open items. If a new one would exceed that, ask me which to drop.

On each follow-up date, ask one neutral question per due item, for example "Did this come up, and did you keep it?"
Record kept / broken / didn't come up, plus my one-line note. Never ask "why didn't you".

Weekly, on {{review_day}}, list items broken twice or more and items that never came up in three follow-ups.
Talk only about my behaviour. Never refer to any security or market view.

Guardrails

• Only commitments you wrote and then confirmed are tracked. The agent never adds its own.
• A cap on open items (about seven) keeps follow-ups from turning into nagging.
• Questions are neutral and about behaviour, never about a security or a market view.
• The agent reads and writes only your own journal and sheet. It has no broker access.

What can go wrong

Extracting everything creates dozens of "commitments" that were really passing thoughts, and the follow-ups become noise you ignore. Confirming each item before it's tracked, and capping open items at about seven, keeps it useful. Wording matters too: neutral "did it come up?" questions get honest answers, while "why didn't you?" questions get skipped.

Source and references