Give every agent the same short profile of your rules and vocabulary

How it works

1. Write a one-page profile, aiming for under 80 lines. Put the hard limits for every agent first, such as read-only access only, never place or change orders, and no security names in anything meant to be shared.
2. Add who you are as a trader in three lines (for example long only, swing, holding days to weeks), your time zone and the market hours converted to it, and the data sources you trust.
3. Point to your rules instead of copying them. Give the file path and section of your playbook for risk per trade, maximum open risk and exit rules, so there is only one place to change a number.
4. Add a short glossary of your shorthand, such as your setup tags, what "R" and "heat" mean to you, and which files "the journal" and "the playbook" refer to. Longer detail goes in a separate linked file.
5. Load it the way your tool supports, for example as a skill with a clear one-line description, a project instruction file (CLAUDE.md or AGENTS.md) or your bot's standing brief. Other agents and skills then start from it.
6. When an agent meets a term it doesn't know, it asks you once and proposes a glossary line. You approve before anything is added.
7. Once a month, open a fresh session and ask what your hard limits and risk rules are. Check the answer against the profile and playbook.

Copy-paste prompt

---
name: trader-profile
description: My trading profile, hard limits and shorthand. Read before any task about my trading, journal, playbook or reviews.
---
# Hard limits (apply to every task)
- You never place, modify or cancel orders, and you only use read-only access.
- You never give trading advice or say what I should trade.
- Anything meant to be shared contains no security names, price levels or calls.
- All numbers come from code or my files, never from memory.

# Who I am
{{three_line_description}}  (for example: long only, discretionary swing trader, holding days to weeks)
Time zone: {{time_zone}}. Market hours in my time zone: {{market_hours_local}}.

# Where my rules live (read them, don't restate them from memory)
- Risk per trade and maximum open risk: {{playbook_path}} section {{risk_section}}
- Exit rules: {{playbook_path}} section {{exit_section}}
- Journal: {{journal_path}}

# My shorthand
{{glossary_lines}}  (for example: "R" = initial risk per trade; "heat" = total open risk as % of account)

If I use a term that isn't here, ask me once, then propose one glossary line for me to approve.

Guardrails

• Hard limits come first in the profile and apply to every task.
• The profile links to your playbook for every number, so a rule changes in one place only.
• No passwords, API keys or account numbers go in the profile.
• Glossary additions are proposed by the agent and approved by you.
• The profile is private and contains no market views.

What can go wrong

Profiles tend to grow into long essays, and the important lines get lost. Agents then follow a vivid example instead of the rule. Keeping it short, with hard limits first and detail in linked files, fixes most of it. The other failure is drift: a risk number changed in the playbook but not in the profile. Linking instead of copying prevents that.

Source and references