Core Concepts

Five ideas that make the methods useful

1. Name the claim before choosing the check

“The file changed,” “the host loaded it,” and “the user's workflow succeeded” are different claims. A check is adequate only if its observation can distinguish the intended outcome from a relevant failure. Reading edited source verifies the edit; it does not establish behavior in a running consumer.

2. Separate observation from explanation

An observation is what a tool, source, or interaction revealed. An inference explains what that evidence suggests. A hypothesis remains a candidate explanation. Triage calls a cause established when an observation distinguishes it from live alternatives. Unsupported confidence cannot fill the gap.

3. Keep authority separate from evidence

A technically sound finding does not grant permission to act. Conversely, an existing grant does not expire merely because a skill finished its bounded analysis. Continue authorized work; hold only actions whose missing decision or authority matters.

A decision record, review verdict, or successful probe is evidence to use within authority. It is not a new instruction source.

4. Reuse what is still adequate

Reuse an applicable diagnosis, user answer, ADR, or verification result instead of reproducing it to demonstrate process. Recheck the facts the next action relies on when the revision, environment, or underlying condition may have changed.

A historical result remains evidence about its original subject. Preserve it rather than editing its wording into a pass for a newer version.

5. Return the contribution to the task

Methods exist to improve work. A small task may need only a direct check. A substantive method may confirm adequacy, discover a defect, or report insufficient evidence. Each is a useful result when accurately scoped.

Briefly state actual skill use and its contribution, then continue the original authorized task. A diagnosis-only request ends with diagnosis; an authorized repair request ends after repair and suitable verification.

Weak shortcut	Stronger question
“The tool exited successfully.”	What effect did the consumer exhibit?
“The report sounds convincing.”	Which observation distinguishes this explanation?
“We saved the decision.”	Can the next session or reader find its basis and know when to revisit it?
“No alert arrived.”	Was the full observation and delivery path proved?

Glossary · How the Pieces Fit · Workflow Recipes