Repeat-Mistake Ledger
One table and four thresholds. The cheapest quality mechanism in the system.
A single table that counts repeated mistakes by class, with fixed escalation thresholds at 1, 2, 3 and 5. The habit that makes it work is logging in the same action as the correction — a ledger filled in during a weekly review pass is abandoned within a month.
What it is
A counter for repeated mistakes, with escalation decided in advance.
# REPEAT LEDGER
## Escalation
| Count | State | Action |
|-------|-----------|-----------------------------------------|
| 1 | log | entry written, nothing else |
| 2 | warning | propose a dedicated rule |
| 3 | systemic | write the rule + assign enforcement L0-L3 |
| 5 | critical | hard failure in review |
## Enforcement levels
L0 prompt text · L1 checklist in a skill
L2 build or hook check · L3 hard gate that fails
## Ledger
| ID | Mistake (class, not incident) | Count | Last | State | Fix |
|----|-------------------------------|-------|------|-------|-----|
| R-001 | [description] | 1 | [date] | log | — |How to use
Write descriptions as classes, not incidents. Concluded absence from a single spelling of a name is reusable; did not find the supplier is not — and if entries never reach count 2, that is almost always why.
Log in the same action as the correction. Not at the end of the session. The moment you notice a mistake is the only moment you have full context on it, and it is also the moment you most want to move on.
At count 3, assign an enforcement level. Writing a rule is not fixing anything — text does not stop repetition on its own, and a ledger full of L0 entries with rising counts is the proof.
What to fill in
Nothing but the ladder, which is already filled. Start empty and let it grow. A ledger seeded with imagined mistakes trains you to look for the wrong things.
License
MIT.