ICC C5 cheat sheet

The five content domains in this blueprint are labeled "unverified" because no published task analysis has been mapped to them; treat any domain weighting as provisional until confirmed.

A cheat sheet entry is only valid if it can be restated as a fact, formula, or decision rule; anything phrased as encouragement is out of scope for this artifact.

Cross-domain questions are resolved by identifying the dominant control objective first, then applying the subordinate domain's rule as a constraint.

Definitions on this exam use the exam's delivery language; standard technical terms remain in English even inside translated stems.

Each flashcard definition must stand alone: no pronoun may refer to the term itself, and no external figure may be required to answer.

When a fact cannot be stated with confidence, the correct authoring behavior is to record null and flag the item for review rather than guess.

Answer stems are self-contained by construction; an option that depends on an unseen exhibit is defective and should be rejected on that ground alone.

Term length is capped at six words and definition length at forty words, so precision of scope beats completeness of coverage per card.

A definition is self-contained when a reader who has never seen the term can apply it correctly to a new case using only the definition text.

Circular definitions repeat the term or a morphological variant; replace the repeated term with its functional role in the process being described.

A six-word term cap forces one concept per card: if a term needs a conjunction, split it into two cards.

Definitions run 15-40 words; below 15 the card usually omits the discriminating condition, above 40 it usually smuggles in a second concept.

Prefer operational phrasing (what is measured, compared, or triggered) over evaluative phrasing (important, critical, best practice).

Standard technical terms stay in English inside non-English definitions; translating them creates false synonyms that mislead retrieval.

Null is a legitimate field value: an unknown date, threshold, or citation is recorded as null rather than approximated.

Term-definition pairs are independently gradable; a card whose answer is inferable from the term's surface form tests vocabulary, not knowledge.

A stem must contain every condition needed to select one option; moving a condition into an exhibit converts the item into a lookup task.

Distractors should encode plausible misconceptions, not random noise; each wrong option must be defensibly wrong for a stated reason.

Negatively worded stems (EXCEPT, NOT) raise misread rates; if used, the negation is capitalized and placed immediately before the target.

Absolute qualifiers (always, never, only) in a correct option usually signal a defective key unless the underlying rule is genuinely absolute.

Numerical options should differ by one order of magnitude or by a common unit error, so the item tests the rule rather than arithmetic.

Option sets of four with one key and three distractors are the default; parallel grammatical form across options prevents form-based guessing.

An item flagged needs_review is excluded from scored pools until a second reviewer confirms or corrects the underlying fact.

Answer leakage occurs when the stem's wording echoes the key's wording; rewrite the stem to use neutral vocabulary shared by all options.

Accuracy outranks volume: a smaller verified set is preferred to a larger set containing unverified claims.

Every factual claim carries a confidence state: verified, provisional, or unknown; unknown maps to null plus needs_review.

Originality is a hard constraint: no item may paraphrase or reference an existing commercial question bank, and provenance is recorded per item.

No claim of official affiliation is permitted; examiner or vendor endorsement must never be implied in stems, rationales, or metadata.

Second-reviewer sign-off is required before a provisional fact is promoted to verified status.

Versioning is per item, not per document, so a single corrected fact does not force re-review of unrelated cards.

A rationale explains why the key is correct and why each distractor fails; rationales without distractor analysis are incomplete.

Recency checks apply to thresholds, dates, and version numbers; a stale figure is treated as unknown until re-verified.

Card counts are distributed across domains by the blueprint, not by author familiarity; a domain with no published weight receives an even share.

Thirty-two cards across five domains yields a baseline of six per domain, with the two remainder cards assigned to the highest-weight domains.

Domain tags are mandatory on every card so coverage can be recomputed without re-reading definitions.

Coverage gaps are detected by counting cards per domain and comparing against the target share, not by counting total cards.

A domain with fewer than six bullets in its cheat sheet signals thin source material and should trigger a content search before authoring.

Duplicate coverage is measured by concept, not by term string; two differently worded cards testing one rule count as one.

Remainder allocation is deterministic: sort domains by weight descending, break ties by domain id ascending, then assign one card each in that order.

Recompute the distribution whenever a card is added, retired, or retagged; the sum must remain exactly thirty-two.