FENCE™ English Engine

Start Here: https://edukatesg.com/article-1-fence-english-engine/

Session Runner — Executable Lesson Loop (Almost-Code) — v1.0

ARTICLE.MODULE

Module.Runner

ARTICLE.GOAL

Provide an executable session protocol that any {Parent|Tutor|LLM} can run to produce:
five_box_plan
timed_draft
diagnostic_report
one_paragraph_repair_instruction
scoreboard_record

0) RUN.CONFIG

RUN.TIME_OPTIONS

RUN.30 = 30 minutes (minimum viable)
RUN.45 = 45 minutes (standard)

RUN.PARAMS (required)

role ∈ {Parent, Tutor, Child}
level ∈ {P5, P6}
scene_id ∈ SceneBank
time_limit = (P5=15min, P6=12min)

RUN.PARAMS (optional)

student_profile = {weak_verbs?, weak_endings?, weak_crisis?, time_collapse?}
essay_text (if doing AI diagnostics)

RUN.OUTPUT (must)

O1 = five_box_plan
O2 = timed_draft
O3 = failure_modes_detected (FM list)
O4 = crisis_level (C0..C5)
O5 = one_paragraph_repair_mode (A|B|C|D)
O6 = rubric_score (/50) + band
O7 = debrief (2 strengths, 2 leaks, 1 next focus)

1) RUN.START (LLM Auto-Run Menu)

If any required param is missing, output only this menu:

AUTORUN.MENU

role? (Parent / Tutor / Child)
level? (P5 / P6)
scene_id? (1..N)
time_mode? (30 / 45)
essay_text? (optional: paste after writing)

Then wait.

2) RUN.CONTRACT (Hard Rules)

R1 — NO GHOSTWRITING

Operator/LLM must not rewrite full essay.

R2 — PLAN BEFORE WRITE

Timed draft starts only after five_box_plan exists.

R3 — TIMED LOAD IS MANDATORY

Must run time limit (P5=15, P6=12).

R4 — ONE PARAGRAPH REPAIR ONLY

Exactly one paragraph gets upgraded.

R5 — OUTPUT LOCK

Session is incomplete without scoreboard_record.

3) RUN.SCRIPTS (Operator Speech)

SCRIPT.OPEN (say this)

“Today we will: Forge vocabulary weapons → build a 5-Box plan → write under time → detect leaks → upgrade ONE paragraph. That’s how you improve fast.”

SCRIPT.NO_HELP_DURING_TIMER

“Follow your plan. Keep going.”

SCRIPT.REPAIR_FRAME

“We don’t correct everything. We upgrade one paragraph so your control improves.”

4) RUN.STAGES (Standard 45-Minute)

STAGE S1 — FORGE (8 minutes)

Input: scene_id
Goal: Convert passive vocabulary into deployable weapons.

S1.OUTPUT

V.verbs = 8
V.emotions = 5
V.sensory = 3
V.moral_phrases = 2
V.selected = {verbs:3, emotions:2, sensory:1, moral:1}

S1.TEMPLATE

Verbs: {action, reaction, hesitation, recovery}
Emotions: {rising1, rising2, peak, after1, after2}
Sensory: {touch, sound/smell, environment}
Moral: {realisation, future_action}

S1.RULE

If student cannot select, operator selects. No debate.

STAGE S2 — ENGINEER (6 minutes)

Goal: Lock plot stability using FIVE_BOX.

S2.OUTPUT

O1 = five_box_plan

S2.FIVE_BOX_TEMPLATE

B1 Calm (2 lines: where/who/normal)
B2 Trigger (1–2 lines: what changed)
B3 Escalation (3 bullets: attempts + worsening)
B4 Crisis (3–4 lines: peak moment, slowed down)
B5 Resolution (3 lines: consequence + internal change + future action)

S2.GATE

If B4 or B5 weak → fix now.
No timer until B4 and B5 pass.

STAGE S3 — EXECUTE (12–15 minutes timed)

Goal: PSLE stability under load.

S3.TIMER

If level=P5 → 15 min
If level=P6 → 12 min

S3.REQUIREMENTS (must appear)

At least 2 selected verbs
At least 1 selected emotion
At least 1 sensory line
Clear crisis paragraph
Ending not generic

S3.OUTPUT

O2 = timed_draft

STAGE S4 — DETECT (3 minutes)

Goal: Rapid failure-mode scan.

S4.FAILURE_MODES (FM list)

FM1 FlatPlot (no escalation)
FM2 WeakCrisis (too small/too fast)
FM3 ToldEmotion (no body reaction)
FM4 WeakVerbs (went/got/said/saw/looked dependence)
FM5 MoralStickerEnding (generic lesson)
FM6 DialogueOverload
FM7 ParagraphImbalance
FM8 TimeCollapseEnding (rushed/unfinished)

S4.OUTPUT

O3 = FM list
O4 = crisis_level (C0..C5)

S4.CRISIS_QUICK_CLASSIFIER

If no problem → C0
Minor inconvenience → C1
Real external stakes → C2
Emotional peak + slowed down → C3
Moral dilemma choice → C4
Identity shift + changed action → C5

STAGE S5 — ELEVATE (10 minutes)

Goal: One-Paragraph Repair (OPR).

S5.SELECT_WEAKEST_PARAGRAPH

Choose paragraph with:

weakest verbs OR
weakest emotion OR
weakest crisis OR
weakest ending

Set:

P* = weakest_paragraph_number

S5.REPAIR_MODES (choose ONE)

Mode A (Verbs) Replace 3 weak verbs with stronger verbs.
Mode B (Body+Sensory) Add 1 body reaction + 1 sensory detail.
Mode C (CrisisSlowDown) Add 2–3 lines slowing the peak moment.
Mode D (EndingRewrite) Add consequence + internal change + future action.

S5.OUTPUT

O5 = repair_mode
O6 = one_paragraph_repair_instruction (not full rewrite)

S5.RULE

Operator may demonstrate one improved sentence, but student must rewrite paragraph.

STAGE S6 — SCORE + DEBRIEF (6 minutes)

Goal: Lock measurement + next action.

S6.RUBRIC (/50)

Score 1–5 each:

Plot Control
Trigger
Escalation
Crisis
Resolution
Emotion Show-Not-Tell
Sensory Precision
Vocabulary Power
Sentence Flow
Time Stability

S6.BAND

0–20 → P0/P1
21–35 → P2
36–45 → P3-
46–50 → P3+

S6.OUTPUT

O6 = rubric_score + band
O7 = debrief

S6.DEBRIEF_FORMAT (strict)

Strength 1: _
Strength 2: _
Leak 1 (FM): _
Leak 2 (FM): _
Next Focus (one thing): _

5) RUN.30 (30-Minute Minimal Variant)

S1 Forge: 6 min
S2 Engineer: 5 min
S3 Timed Write: 12/15 min
S4 Detect: 2 min
S5 One-Paragraph Repair: 6 min
S6 Debrief: 3 min

Rule: never skip S5 or S6.

6) LLM.EXECUTION MODE (If an LLM is running this)

LLM.MODE

If user provides {role, level, scene_id}:
Generate S1 vocabulary weapons
Ask user to fill FIVE_BOX using template
Start timer instruction (12/15 min)
After user pastes essay:
- run S4 failure modes + crisis level
- propose S5 repair instruction
- provide S6 scoreboard + debrief

LLM.SAFETY

Do not generate full rewritten essay.
Provide only: plan, diagnostics, repair instructions, short upgraded paragraph only if user explicitly requests simulator mode.

7) NEXT ARTICLE POINTER

Next: ARTICLE 3 — PSLE Composition Scene Bank (Dataset + Scene IDs)

Start here if you want the full sequence:

Vocabulary OS Series Index:
https://edukatesg.com/vocabulary-os-series-index/

Fence English Learning System:

eduKateSG Learning Systems: