03 · Journal
Rule-break review
Cluster the week’s broken rules — size, time, revenge entries — and write the one change you will actually enforce.
- Level
- Intermediate
- Cadence
- Weekly, after fills are journaled
- Read
- 7 min
- Updated
Definition
A rule-break review classifies trades against a rule list that existed before the week. Profit does not erase a break. The output is one change to enforcement, not a personality critique.
Who it is for
Traders who already have written rules and suspect they are negotiating with them in real time.
Job to be done
Compare the week’s fills to rules you wrote in advance and choose one enforcement change.
Human gate. You pick the one change. The agent may list breaks. It may not add new rules during the review.
Inputs and outputs
Inputs
- The rule list as it stood on Monday: max size, session window, add rules, stop rules
- Journal entries or fills for the week
- A note of any rule you changed mid-week, with the time you changed it
Outputs
- Each trade tagged followed, broken, or unclear
- The rule that broke most often
- One enforcement change you can apply next week
- No new discretionary rules invented in the same sitting
Step-by-step workflow
- 01Freeze the rule textPaste the rules as they were at the start of the week. Do not tidy them first.
- 02Tag every intentFollowed, broken, or unclear. Unclear means the rule was ambiguous, which is itself a finding.
- 03Count, do not narrateWhich rule, how many times, and whether the broken trades made or lost money. Both counts stay.
- 04Pick one breakThe most frequent, or the one that can hurt the account. Ignore the rest until next week.
- 05Write an enforcement changeA smaller max, a session alarm, or a checklist you must fill before entry. Not a promise to "be better".
- 06Human gateYou commit to that one change in the journal. The agent does not email you motivation.
Suggested tool types
These are categories of tools, not product picks. Crypto Agent Desk does not certify vendors and does not run an affiliate program.
| Type | Why it is here |
|---|---|
| Journal file from the fill workflow | The review needs intent ids, not a fresh story. |
| Chat model | Tags trades against the written list and counts. It should quote the rule. |
| Spreadsheet | Counts across weeks show whether the same rule keeps breaking. |
Risks and failure modes
Rules edited by the outcome
Adding "except when it worked" during the review deletes the rule. Mid-week edits must be timestamped, not silent.
Five changes
A list of improvements is how nothing is enforced. The playbook allows one change.
Moral language
Discipline lectures are not actionable. The tag is followed or broken, plus the next constraint.
Operator notes
The useful output is embarrassingly small: one rule, one count, one change. Everything else is how you avoid the change.
Tag set
Followed. Broken. Unclear. There is no tag called “broken but smart”.
Prompt you can paste
Here are my rules as of WEEK_START: RULES.
Here are journal entries: ENTRIES.
Tag each intent followed, broken, or unclear. Quote the rule you used.
Count breaks by rule. Note P&L beside the count but do not let P&L change the tag.
Recommend no more than one enforcement change, and phrase it as a constraint.
Do not add a new trading idea.Done when
Next week’s rule list differs by one sentence you can point to, or you explicitly kept the list and the break is still visible.
FAQ
What if a broken rule made money?
It is still broken. Record the P&L so you can see the temptation. Do not let the P&L retcon the tag.
What if I had no written rules?
Then this review cannot run. Write the rules first and review next week. Do not let the agent infer rules from your winners.
Can unclear tags become new rules immediately?
Only if you are replacing an ambiguous sentence with a precise one. That counts as your one change.
Published . Updated . This is an operating playbook, not a recommendation to buy, sell, or hold any asset.