Page 75 - CyberSecurityEssentialsEnglish

Source: use the report only.
Structure: the headline finding, three indicators, two risks, two possible decisions, and
questions about missing data.
Rule: carry the figures across exactly as written, link each figure to its page, and separate
what the report states from your own inference.
Tone: executive and neutral.

Verification: at the end, add a short list of the claims that need human review.

The strength of this template does not come from its length. It comes from the fact that it

defines the function of the summary, its source, the order of information, the policy on
figures, and the separation between text and inference. If the report is simple, some
elements can be dropped. If the decision is high-impact, a summary alone is not enough and

the original pages must be consulted.

▣ Practical example: a graduated teaching explanation template
Task: explain the concept of cash flow to someone running a small business who has not

studied accounting.
Structure: a definition in two sentences, a simple numerical example, a common mistake,
a comprehension question, then a short application.
Language: simple standard prose, with the technical term given at first appearance.
Constraint: do not confuse cash flow with profit, and if you need to assume a figure, say
so clearly.

Success criterion: the reader can explain how a business can be profitable and still run
short of cash.

This template does not merely ask for a simple explanation. It specifies the prior knowledge,
the conceptual distinction that must be preserved, and evidence that understanding has

occurred. The comprehension question and the application prevent a text that is easy to
read but that builds no real understanding.

▣ Practical example: a structured extraction template
Task: extract the following fields from the complaints: category, product, date, urgency,
required action, and a supporting quotation.

Missing values: use null and do not guess.
Categories: use the approved list only.
Format: one row per complaint, with a unique identifier. Verification: every non-empty
value must have a supporting quotation. Unclear
cases: set needs_review to true when a complaint could fall into two categories.
In extraction, traceability matters more than style. A supporting quotation, a policy for
missing values, and a state for review together reduce the temptation to fill fields. If the

output will move into another system, a schema and programmatic validation must be
added, as a later section explains.

74