Z2 is where students “can write sentences” but still lose marks because the paragraph has no spine, drifts, or doesn’t answer the task. This is the paragraph-as-machine layer.

Z2.0 Canonical Paragraph Object (PO)

ID: ENG.Z2.PO.v0_1

			
PO = {
  TaskType,          # NARRATE | EXPLAIN | ARGUE | DESCRIBE | REFLECT
  ClaimScope,        # what this paragraph is responsible for
  Spine,             # C → R → E → X → L
  Cohesion,          # reference chains + connector routing
  EvidenceMode,      # example | scene | fact | quote | statistic | analogy
  Control,           # drift control + relevance checks
  Finish             # closure + bridge to next paragraph
}

		

Z2.1 Paragraph Spine (LOCK)

ID: ENG.Z2.SPINE.CREXL.v0_1

Canonical spine:

			
C = Claim / Point (what I’m saying)
R = Reason (why it’s true)
E = Evidence / Example (show it)
X = Explanation (connect evidence to claim)
L = Link / Loop-close (wrap + point to next)

		

Most weak paragraphs are missing R or X.
They “tell” and “example” but never explain.

Z2.2 Node Directory (Paragraph Moves)

Z2.NODE.CLAIM

AtomID: ENG.Z2.NODE.CLAIM

Role: sets point + scope
CommonLeaks: vague claim; too many points; claim not tied to prompt
Sensors: ClaimClarityScore, PromptAlignment
Repairs: 1-sentence claim rewrite × 10

Z2.NODE.REASON

AtomID: ENG.Z2.NODE.REASON

Role: creates causality (because-chain)
CommonLeaks: reason absent; “because” used but no logic
Sensors: ReasonPresenceRate
Repairs: Because-chain drill: 3 layers deep

Z2.NODE.EVIDENCE

AtomID: ENG.Z2.NODE.EVIDENCE

Role: supplies concrete support
EvidenceMode enum: {EXAMPLE,SCENE,FACT,QUOTE,STAT,ANALOGY}
CommonLeaks: evidence too general; irrelevant evidence
Sensors: EvidenceSpecificityIndex
Repairs: Zoom-in drill: add 3 sensory/observable details

Z2.NODE.EXPLANATION

AtomID: ENG.Z2.NODE.EXPLAIN

Role: converts evidence into marks
CommonLeaks: “example dump”; explanation missing
Sensors: ExplainAfterEvidenceRate
Repairs: So-what drill: “This shows…” sentence forced

Z2.NODE.LINK_CLOSE

AtomID: ENG.Z2.NODE.LINK

Role: closes loop; prevents drift; bridges next paragraph
CommonLeaks: abrupt ending; repeats claim only
Sensors: ClosurePresenceRate
Repairs: 1-line closure + next hook

Z2.3 Cohesion & Flow (Edges)

ID: ENG.Z2.COHESION.v0_1

Z2.EDGE.REFERENCE_CHAIN

Role: pronoun clarity + noun repetition balance
CommonLeaks: “he/she/they” unclear; sudden new noun
Sensors: ReferenceBreakCount
Repairs: anchor-noun rule: re-name noun every 2–3 sentences

Z2.EDGE.CONNECTOR_ROUTING

Role: logic connectors match actual logic
Core connectors by function:
Cause: because, therefore, so
Contrast: but, however, although
Addition: and, moreover
Sequence: then, next, finally
Example: for example, for instance
Sensors: ConnectorCorrectnessScore
Repairs: logic swap drill (rewrite with correct connector)

Z2.EDGE.SENTENCE_VARIETY

Role: avoid choppy / repetitive structure
Sensors: VarietyIndex (simple vs complex sentence mix)
Repairs: combine 2 simples → 1 complex drill

Z2.4 Drift Control (Prompt Alignment)

ID: ENG.Z2.DRIFT_CONTROL.v0_1

Z2.SENSOR.DRIFT

Definition: % of sentences not serving the paragraph claim
Simple test: each sentence must answer “How does this support C?”
Sensors:
DriftRate
OffPromptCount
Repairs:
Underline-task-words
C-sentence must include 1 task word
Last sentence must loop back to task

Z2.5 Paragraph Types (TaskType Templates)

1) Narrative Paragraph (composition)

ID: ENG.Z2.TPL.NARRATE.v0_1

			
C = what happens in this beat
R = why it matters (emotion / intention / stakes)
E = scene details (actions + senses)
X = inner thought / consequence
L = hook next beat

		

Leak patterns:

too much telling, no scene
scene but no “why it matters”

2) Explanatory Paragraph (comprehension / situational writing)

ID: ENG.Z2.TPL.EXPLAIN.v0_1

			
C = main explanation point
R = causal reason
E = example / simple illustration
X = connect back (“This is why…”)
L = transition

		

3) Argument Paragraph (discursive)

ID: ENG.Z2.TPL.ARGUE.v0_1

			
C = claim (position)
R = reasoning (principle)
E = evidence (fact/example)
X = warrant (“Therefore…”)
L = tie to thesis / next claim

		

4) Descriptive Paragraph

ID: ENG.Z2.TPL.DESCRIBE.v0_1

			
C = what is being described + angle
E = 3–5 sensory details grouped
X = effect/meaning (mood/impact)
L = close image / shift focus

Z2.6 Exam Compatibility Layer (PSLE / Secondary)

ID: ENG.Z2.EXAM_LAYER.v0_1

PSLE Composition (marks logic)

rewards: clear events + emotion + coherence + vocabulary control
highest ROI: Narrative template + drift control + connectors

Secondary (argument/inference)

rewards: reasoning + explanation + register control
highest ROI: Argue/Explain template + warrant sentences

Z2.7 Z2 Sensor Pack (Instrument Panel)

ID: ENG.Z2.SENSORS.v0_1

Measured per paragraph:

P2.S1.SpineCompleteness (C,R,E,X,L present?)
P2.S2.ReasonPresenceRate
P2.S3.ExplainAfterEvidenceRate
P2.S4.EvidenceSpecificityIndex
P2.S5.ReferenceBreakCount
P2.S6.ConnectorCorrectnessScore
P2.S7.DriftRate
P2.S8.VarietyIndex

Target sequence:

Spine completeness
Drift rate
Explanation after evidence
Cohesion (reference/connectors)

Z2.8 Failure Tags (Repair Router)

ID: ENG.Z2.FAILTAGS.v0_1

FT.Z2.NO_CLAIM
FT.Z2.NO_REASON
FT.Z2.EXAMPLE_DUMP
FT.Z2.NO_EXPLAIN
FT.Z2.WEAK_EVIDENCE
FT.Z2.DRIFT
FT.Z2.CONNECTOR_MISMATCH
FT.Z2.REFERENCE_BREAK
FT.Z2.CHOPPY

Z2.9 Repair Loops (10-minute drills)

ID: ENG.Z2.REPAIRS.v0_1

RL.Z2.C10 = write 10 claims from prompts (1 sentence each)
RL.Z2.BECAUSE3 = claim → because → because → because
RL.Z2.SHOW3 = replace 3 “tells” with scene details
RL.Z2.SOWHAT = force “This shows…” after each example
RL.Z2.DRIFT_LOCK = highlight each sentence: supports C? yes/no
RL.Z2.CONN_SWAP = rewrite 8 lines with correct connector
RL.Z2.REF_CHAIN = mark pronouns → add anchor nouns
RL.Z2.COMBINE = combine 6 sentence pairs into complex sentences

Copy-Paste WordPress Spec Block (Z2)

ENG_OS.Z2.v0_1

Paragraph = PO(TaskType, ClaimScope, Spine C→R→E→X→L, Cohesion, EvidenceMode, DriftControl)
Canonical spine: Claim, Reason, Evidence, Explain, Link
Sensors: spine completeness, drift rate, explain-after-evidence, cohesion (reference/connectors), specificity, variety
Templates: Narrate / Explain / Argue / Describe
FailTags: NO_REASON, EXAMPLE_DUMP, NO_EXPLAIN, DRIFT, CONNECTOR_MISMATCH, REFERENCE_BREAK
Repairs: because-chain, so-what, drift lock, connector swap, show-not-tell

ENGLISH OS v0.1 — Z2 Paragraph Engine Directory (Almost-Code)