Classify errors as execution, knowledge, or judgment failures before correcting — each type needs a fundamentally different fix

Why This Is a Rule

The default response to any error is "try harder" or "be more careful" — which addresses execution failures but does nothing for knowledge or judgment failures. Classifying before correcting prevents mismatched interventions.

Execution failure (wrong doing): you knew what to do but didn't do it correctly. You forgot a step in a checklist, made a typo, or rushed through a procedure. The knowledge and judgment were right; the implementation failed. Knowledge failure (missing information): you didn't know something critical that would have changed your action. You used an outdated API, didn't know about a regulatory requirement, or lacked a key piece of context. Judgment failure (incorrect assessment): you had the relevant information but weighed it incorrectly, drew the wrong inference, or misjudged probabilities. You knew the risks but underestimated their likelihood.

Each type requires a fundamentally different correction (Match correction type to error type: procedural for execution, epistemic for knowledge, calibrational for judgment). Applying the wrong correction — like adding a checklist for a judgment failure — creates the illusion of fixing the problem while leaving the actual cause untouched.

When This Fires

• After any error, failure, or unexpected negative outcome — before designing any fix
• During post-mortems and retrospectives when analyzing what went wrong
• When the same category of error recurs despite corrections — the classification may be wrong
• Complements Diagnose before redesigning — identify whether trigger, condition, or action broke before changing anything (agent component diagnosis) with a broader error taxonomy

Common Failure Mode

Defaulting to "execution failure" because it's the easiest to fix: "We need to be more careful" or "We should double-check our work." If the error was actually a knowledge failure (you didn't know about a new requirement), more careful execution of the old process won't help — you'll carefully execute the wrong thing. If it was a judgment failure (you misjudged risk), checklists won't help — you'll check all the boxes and still make the same misjudgment.

The Protocol

(1) After any error, ask three diagnostic questions: Execution check: "Did I know what the right action was?" If no → it's not an execution failure. If yes → "Did I perform the action I intended?" If no → execution failure. Knowledge check: "Was there information I didn't have that would have changed my action?" If yes → knowledge failure. What information was missing? Where could I have found it? Judgment check: "Did I have the relevant information but draw the wrong conclusion or misjudge probabilities?" If yes → judgment failure. What distorted my assessment? (2) Classify as one type (occasionally two overlap). (3) Apply the type-matched correction (Match correction type to error type: procedural for execution, epistemic for knowledge, calibrational for judgment).