A plan is three things: the questions the operator must answer, the look-and-try steps with the question each one asks, and the measurements — each taken at every point the workbook names, with the one limit that judges it.
The rules drawn as the tree they form. A box asks one question, a branch
carries one answer to it, and every path ends in the whole plan a device with
those answers receives — visual check, functional check and every measurement, in the
order the workbook gives them. The questions are split into the answers the model lists,
so one answer is one branch and a path is one kind of device. The last question is the
operator's own: how he connects the device — one branch per button his
instrument offers for that sequence, as the operator_actions_at_start sheet
states them. That choice adds and removes nothing from the plan; it decides which
measurements he can reach, and one he cannot reach is struck through. Point at
any step and the panel names the branch that put it there, the rule row it comes from,
and which button does reach it. Drag the small map to move around; nothing here can be
edited.
The same rules, reshaped so that every level narrows: one question, one
branch taken, and the leaf holds the whole sequence. This is what an operator does with
the appliance in hand. His rule table cannot be read this way — its branches overlap,
because a rule marked ANY fires whatever the answer is and "not NONE" also admits an
unbonded device — so a device travels down several of its branches at once. Splitting
each question into its own values gives 32 cells; merging the ones whose plan is
identical leaves 29. Every one of them was walked against the Prolog engine, one
question at a time, and the two agree on all 31 488 devices. The last question is the
operator's own — how he connects the device — which splits those 29 into
51 leaves: the steps are the same under every button, because no rule reads the
connection, and what changes is which measurements that button can reach. One it cannot
reach is struck through.
Every sequence the workbook states, with its rules. A step hangs under its sequence; the chips on it are the conditions the rule states, and each condition says how that field is answered on that sequence — required of the operator, started at a stated value, or with no default at all. A step with no chip always runs. The same rules are drawn as a tree in the first tab; this list is the one to read when you want the limits and the dwell windows beside each step.
Pick a device and answer as an operator would, then choose how the device is connected — both ends in the instrument, or one end loose and the probe. The hookup decides which measurements can be taken at all, so the plan is built per pick, and a measurement the hookup cannot reach says so instead of vanishing. Every measurement shows one row per point it is taken against, with the limit that judges that point. Clearing a field shows what the workbook does with an empty answer: either the field's stated default steps in, or the plan is refused and the missing field is named. Nothing is guessed.
Every field, on every sequence: whether the operator is asked for it, and what its answer starts at. NONE means there is no default — said as a value rather than by leaving a cell empty. Nothing here is read out of prose.
His own electrical matrix — twenty lines written by hand, independent of the rules, saying which measurements each sequence runs and how many times. It is the evidence the rules are checked against, and the Class I block is a complete enumeration: three conductive-parts cases times two insulation-skip cases.