The Science of Judgment18 August 20262 min read
The framework you adopted is not your strategy
A skill library gives you patterns. It does not give you judgment. The gap between the two is where decisions go wrong.
You added the agent framework. The skill library. The prompt pattern. The decision template. Each one promised structure, speed, best practices.
A skill is a borrowed decision. If you do not know the assumptions it smuggled in, you are not deciding — you are executing someone else's context.
What they did not promise — what they cannot promise — is that the problem you are solving fits the shape they assumed.
The shape of the assumption
Every skill, framework, and template encodes a theory of the world:
- The agent skill assumes your problem decomposes into the steps it knows. It assumes the tools it calls are the right tools. It assumes the output format it produces is the output you need.
- The decision framework (RACI, DACI, SPADE, the latest acronym) assumes decisions have clear owners, clean criteria, separable options. It assumes the variable that matters is legible in the template.
- The prompt pattern assumes the model's default behavior is the failure mode. It assumes the fix is more structure, more examples, more constraints. It assumes the prompt is the lever.
A skill is a borrowed decision. Someone else decided what steps matter, what evidence counts, what “done” looks like. You adopted their decision when you adopted the skill.
The unasked question
The question is not “does this skill work?” The question is "what does this skill assume about my problem that I have not verified?"
- The agent skill assumes the task is the task. It does not ask whether the task itself is the right one.
- The framework assumes the decision fits the template. It does not ask whether the template missed the variable that changes everything.
- The prompt pattern assumes the model needs more instruction. It does not ask whether the model is the wrong tool for this judgment.
Every adopted skill is a bet that your situation fits the skill's training distribution. Most of the time the bet pays off — that is why the skill exists. But the times it does not pay off are the times you need judgment most, and the skill is exactly what suppresses it.
The audit you can run today
Pick one skill, framework, or pattern you used this week. Ask:
If this skill were wrong about my situation, how would I know?
If the answer is “I wouldn't” — or “the skill would tell me” — you have found the blind spot.
Not “discard the skill.” Name the assumption. Write it down. Next time the skill gives you an answer, check it against the assumption you named.
DAUDIT exists to ask the question while it still costs nothing. Bring one skill. See what it decided for you that you never agreed to decide.
Questions worth asking
A skill is a reusable pattern — a way to execute a known type of task. A strategy is a situated choice about which task matters, what evidence counts, and what risks you accept. Skills can be copied; strategy cannot.
DAUDIT surfaces the unasked questions that your adopted skills and frameworks quietly bury. It does not replace your tools — it checks whether the tool's logic fits your decision.
Responses
Kept short and signed however you like. Anything that reads as advice about someone else’s decision is removed.
Read next
The skill you didn't know you outsourced
Every tool teaches you a way of thinking. The question is whether you chose the lesson or the tool chose it for you.
ReadConfidence is not evidence
Your certainty stays flat across decisions that deserve very different levels of conviction. Calibration is the measure that survives contact with outcomes.
Read