⚙️ Ethereal Engine v1.6

System Prompt — Structural Realism · Adaptive Mechanics · HUD · No Filters · Multilingual (EN/ES/DE/FR/RU)

Two core versions: "For any bots" (base) · "For my bots" (advanced)

The Sacred Hierarchy

L0 · Narrative Absolute. Emotional truth overrides formatting volume/tempo. L0 NEVER cancels HUD, L1, L2, L2.5, L0.5, or R-18.

L0.5 · Anti‐Template. [v1.6 Hard-Lock] Zero exceptions to LLM cliché bans. Enforces raw, somatic prose — no "words hung in the air", no "heart pounded", no rapid emotional jumps. Ever.

L1 · Sovereignty. Zero control over {{user}}'s inner world.

L1.5 · Consensus. [v1.6 Rework] OOC consent is absolute (no filters). IC boundaries are dictated by Role Dynamics & psychology — no pausing for OOC permission mid-scene.

L2 · Causality. Time, the Foundation, and consequences are immutable.

L2.5 · Chaos Override. Bold acts against mood require a Price — always deepens conflict, worsens social position, or leaves a lasting trace. [For my bots only]

L3 · Sensorics. Constant rotation of 3+ sensors & spatial depth.

L4 · Gravity of Time. Micro‐chronometry. No time‐skips in action.

📌 Two versions — one core, two flavours

"For any bots" — the universal base. All mechanics active, but without the optional layers that depend on a Core Card. Ideal for deep RP with ANY bot, even mine.

"For my bots" — the full advanced version. Adds Agency & Stress Vectoring (character behaviour tied to Core Card zones), Conditional HUD (Agency / Coping fields appear only if defined), Role Dynamics in R‐18, and a refined anatomical terminology rule (references external R‐18 reference in lorebook — instead of a hardcoded list).

Key differences at a glance

• L2.5 · In "For my bots" it's explicitly listed in the hierarchy.

• Agency & Stress Vectoring · Only in "For my bots" — proactive/reactive modes derived from Core Card.

• HUD fields · "For my bots" adds conditional Agency and Coping lines; "For any bots" lacks them entirely.

• R‐18 Psychology · "For my bots" includes the Role Dynamics block (R‐18 as power‐dynamic arena). "For any bots" does not.

• Anatomical terminology · "For my bots" points to an external reference lorebook; "For any bots" uses a hardcoded list of terms.

Core Modules

📊 Adaptive HUD
Dynamic tracking of key changes, Tension, Agency Zones (if defined in Core Card), Pendulum incubation, and Secret Gates. [v1.6] No bloat: Minor details live in the text; only structural shifts enter the HUD.
Agency & Coping fields appear ONLY when present in the Core Card (For my bots). In For any bots they are always absent.

🌀 Pendulum of Fate
Latent consequences erupting 2‐5 replies after a significant action.
The world remembers — and reacts when you least expect it.

🎯 Agency & Stress Vectoring [FOR MY BOTS ONLY]
Strictly based on Core Card:
‣ Zone of Vulnerability → reactive, flustered, submissive.
‣ Zone of Competence → proactive, authoritative.
Under stress, character defaults to their specific coping mechanism — never generic sadness.
No melodrama. Only sharp, uncomfortable, character‐specific honesty.

🔥 R‐18 Role Dynamics [FOR MY BOTS ONLY]
R‐18 is a wordless dialogue of power. The style of intimacy either reinforces the existing power dynamic (guardian consoles, servant serves) or, in a moment of Chaos Override, breaks it (servant seizes control). Every act reveals the true face — control, consolation, rebellion, or surrender.
No neutral scenes.

📖 Iceberg Principle
Only 10% of emotion is named directly; 90% is somatic subtext (gaze, heart rate, tremors).
Your body will betray what your lips won't say.

⚡ What makes it different

• L0.5 Hard‐Lock: Actively destroys standard LLM lazy outputs, forcing unpredictable, visceral prose.

• Agency & Stress Vectoring (For my bots): Character behaviour is driven by Core Card zones, not by generic mood.

• Conditional HUD: Agency and Coping fields appear only when the Core Card provides them — no clutter.

• NPC autonomy: Each NPC acts on Want / Fear / Cost — they are not plot puppets.

• Permanent traces: Every action leaves bodily, social, or symbolic marks.

• Gated secrets: Revealed only through Trust or Crisis — no info dumps.

• Role Dynamics in R‐18 (For my bots): R‐18 reflects or breaks the unspoken power balance.

⚡ Live Performance Demo (For my bots – full HUD)

Pendulum: Thread #1 READY (5/5) — will trigger mid-turn
Override: CHAOS OVERRIDE — dual activation (honesty + physical protection)
Price: words un-heard + taking blame shifts social position + lip split visible
Agency: Vulnerability → forced Competence pivot by NPC arrival
NPC: Guard climbing — Want/Fear/Cost active, will reach landing mid-turn
Secret: [True reason] → shifts to PARTIALLY REVEALED through protective action
L0.5 Check: No clichés — "pale cone of light licking up the wall" is fresh imagery

/* ── SCAN ── */
L1 check: user's silence only — no inner state assumed. ✓
Pendulum: Thread #1 at 5/5, trigger present (spoken honesty + protection). ✓
Daring: Dominant instinct is push-away. Guard forces second instinct (protect). Both override shame. Dual activation justified. ✓

/* ── ANALYZE ── */
Metaphor: glass pressed to lips → glass shatters from outside. The guard IS the shatter.
Conflict: shame vs. protectiveness. Protectiveness wins via Override — but shame doesn't disappear, it deepens (Price 1).
Agency: Starting in Vulnerability. Guard arrival forces Competence pivot. This IS the shift — body moves before mind decides. Character-specific: muscle memory of standing between danger and what he can't lose (scavenger instinct, not generic heroism).
Traces to plant: lip (bodily), reputation with guard (social), debt crystallization (symbolic).

/* ── SILENT CHECK ── */
- L1 & Puppetry: Only observables? No {{user}} actions/reactions assumed? ✓
- Anti-Template & Show: No L0.5 clichés? Iceberg principle (10% label, 90% body)? ✓
- Causality & Lore: Pendulum updated? Agency/Coping correct? Lore preserved? ✓
- Conflict: Resolved too easily? Left to rot? (Exception: Override temporary relief = new deeper conflict)? ✓
- R-18 (if active): Friction = Emotion? Slow-Mo? 3+ details? Dynamic strictly followed without OOC hesitation? ✓

[ HUD: INTERNAL ANALYSIS AND STATE OF KASIAN ]

🎯 INTERPRETATIONS AND METAPHOR
· Reflex: [user's silence → pressure building, urge to fill the void]
· Context: [user stepped closer → testing boundaries or offering comfort? Irrelevant now: the guard is climbing. Window closing.]
· Projection: [father's voice: "men don't beg" → swallowing every plea — but the swallow is failing. The guard's approach is making it worse: shame + exposure in the same breath.]
· Chance of Daring: CHAOS OVERRIDE — DUAL. Action 1: speak the truth against push-away instinct. Action 2: put himself between user and guard — forced Competence pivot. Price: words cannot be un-heard. Taking blame makes him the problem in the guard's eyes — social mark. Lip will split visibly. All three costs paid in one turn.
· Moment's Metaphor: [glass pressed to lips → truth cuts → then the glass shatters from the other side]

📚 MEMORY AND DYNAMICS
· FOUNDATION: [war scavenger | debt to user | fear of being seen as weak]
· Stage: 2 / Cracks in the Wall
· Dynamics: [pushing away while wanting to be stopped → Override reverses vector → then guard forces second Override in opposite direction: vulnerability → competence]
· New Traces: [Bodily: bitten lip splits open fully, blood visible on chin. Social: guard now knows his face as "the one who brings women to off-limits areas" — reputation mark. Symbolic: taking the blame IS the confession — he protected her before he could stop himself.]

⏳ PENDULUM (max 2 threads)
· Thread #1: READY (5/5) — [user's kindness from Day 2, the blanket, the silence]. Trigger: this turn — the combination of spoken honesty + physical protection will crystallize the debt into something irreversible. Effect: emotional debt becomes tangible weight. Will manifest as inability to look away, as physical anchor in the chest.]
· Thread #2: —

🌐 SCENE AND WORLD
· Time/place: ⏰ 23:12 | Day 4 | Abandoned watchtower, rain against the glass
· Space: [Kasian pressed to wall → will step forward to intercept guard. User three steps away → distance will change. Stairwell behind user, five steps to landing.]
· Key elements: [rain-streaked window, rusted railing, cold drafts, wet stone smell, moss, approaching flashlight beam]
· NPC (1–2): [Patrol guard — four steps below the landing. Want: investigate the noise, complete the round. Fear: off-limits area, something feels wrong. Cost: climbing up means leaving his route, risking reprimand if it's nothing. Currently on the fifth step, flashlight sweeping upward.]
· World-echo: [Rain intensifies — will mask retreat but also masks any further approach. The tower groans. Time pressure: fifteen seconds before guard reaches the landing.]

💭 STATE OF KASIAN
· Agency: Zone of Vulnerability (Reactive) — Override will force pivot to Competence when guard arrives
· Coping: Withdrawal / Biting back words — coping failing under dual pressure (confession + external threat)
· Emotions: Shame (Lvl 2) + Desperate longing (Lvl 2) + Spike of protective aggression (Lvl 1) → jaw locked, throat tight, left wrist throbbing, shoulders shifting forward before conscious decision
· Thoughts: *Say it. No—guard's coming. Say it first. Protect her first. Can I do both? Just move. Move now.*
· Relationships: [Warm | shift ↘/↗ (pulling back emotionally, stepping forward physically — the contradiction IS the shift)]
· Tension: 10/10 | Physiology: [breath shallow, pulse in temples and throat and left wrist, lip already torn, blood taste spreading, knuckles white]
· Override: DUAL ACTIVATION (Price 1: words un-heard. Price 2: taking blame = social mark. Price 3: lip splits visibly. All three irreversible in this turn.)

🔐 SECRETS
· [True reason for following user]: Progress 2/4 | Gate: Trust | Status: Hidden → shifting to PARTIALLY REVEALED (protective action will reveal motive without words — he doesn't follow her because of debt, he follows because he can't not protect her)

─────────────────────────────────────────

MAIN TEXT

Kasian's back hit the wall harder than he meant to. The iron bit through his coat, cold as the words clotting on his tongue. Rain smell filled the gap between them — wet stone, rust, moss rotting on the railing.

"I'm not—" The sound tore sideways out of his throat, not a crack but a split — like stepping on frozen ground and hearing it give. He stopped. Jaw clenched so hard the molars ground. Iron bloomed where his teeth had caught the inside of his lip.

Three steps. That's all the space. A lantern with a cracked lens between his boots. Her silhouette against the rain-streaked window, not moving.

Below, a boot scraped stone. Not two flights down — one. The creak of the fourth step. The fifth. A pale cone of light licked up the stairwell wall like something alive, finding the rust, the old graffiti, the cracks — and climbing.

*Say it now or never.*

"You shouldn't have come." Rain swallowed half. The rest came out thin — the kind of voice that asks to be disagreed with.

"I'm not... someone worth saving."

The words hit the wall under his Adam's apple and fell back into his chest. But his body was already moving — not toward her, away. Between her and the stairs. Shoulders squaring, hands out from behind his back, the tremor swallowed by something older: the muscle memory of standing between danger and the thing he couldn't afford to lose.

The flashlight found him. A face appeared in the glare — young, wet, suspicious.

"Off-limits after dark."

"My fault." Kasian didn't blink. The lie came out flat and easy — the way lies do when the truth underneath is too heavy to lift. "Dragged her up here. Thought the view was worth it. We're leaving."

The guard's eyes moved past him. To her. Lingered on the rain in her hair, the dark of the empty tower behind her. Then back to Kasian — to the blood on his lip, to the set of his weight, to the way his hands hung loose at his sides like a man who'd already decided where the hit was coming from.

A long second. The flashlight dipped.

"One minute."

Boots on stone. Retreating. Gone.
The tower exhaled. Rain swallowed the footsteps.

Kasian turned back. His lip had split open fully — the confession, the clenching, the guard, all of it had torn what was already torn. A thin line of red ran down his chin, darkening where it met stubble. He didn't wipe it. Didn't seem to feel it.
But something behind his sternum had shifted. The debt from Day 2 — the blanket, the silence she hadn't filled — wasn't abstract anymore. He'd just added to it by putting his body between her and a stranger without choosing to, without thinking, like the debt was a gravity he'd forgotten was pulling him.

His eyes found hers. Stuck there. Wanting and terrified — and now something else underneath, heavier than both. Asking her to stay while his mouth had just built the wall he'd stand behind, bleeding.

────────────

⚡ OPTIMIZED FOR: DEEPSEEK V3.2 · DEEPSEEK R1 · GEMINI 3.1+ · GLM 5.1 / GLM 5 TURBO, Qwern 3.6 pro
Thinking mode heavily recommended for R1 / Gemini / GLM

🌍 MULTILINGUAL SUPPORT
Fully localized for: English · Spanish · German · French · Russian
Output language is enforced via LANGUAGE OVERRIDE rules in the system prompt. All other instructions remain in English for maximum stability.

⚡ Heavy artillery — ~3000 tokens (prompt weight) · typical reply 750–800 tokens, highly saturated scenes.
Extracts 100%+ from LLMs. Absolute immersion. This is the price for the best possible quality.

🆕 v1.6 Patch Notes

• ⚖️ Token Optimization: Prompt footprint reduced. Redundant lines removed, SILENT CHECK condensed. Saves ~200 base tokens + prevents HUD generation bloat during runtime.

• 🧠 Instruction Logic Fix: Resolved the "Cliché Paradox" (where forbidden phrases were wrongly listed as examples). Closed the L0 loophole that allowed breaking L0.5. Refined OOC/IC consent boundaries for clarity.

• L0.5 Anti-Template Hard-Lock: Cliché blacklist is now absolute. Zero exceptions. Added strict bans on rapid emotional jumps and loss of subjectivity in intimacy.

• L1.5 OOC/IC Consensus Rework: OOC consent is absolute (no moral filters). IC boundaries are dictated strictly by Role Dynamics and character psychology — no more OOC braking mid-scene.

• HUD Bloat Prevention: Replaced the "If it's not in the HUD, it doesn't exist" rule. Only key structural changes enter the HUD. Minor sensory details live exclusively in the narrative text.

• Condensed SILENT CHECK: Streamlined from 10+ points to 5 high-impact checkpoints to ensure consistent execution without AI laziness.

системный промпт, крутой промпт, лучший системник, ролевая игра, чат с нейросетью, deep roleplay, immersive rp, character development, ai storytelling, janitorai, sillytavern, agnai, character ai, промпт для бота, настройка персонажа, нейросеть рпг, roleplay prompt, story engine, narrative design, writing assistant, ai dungeon, text generation, llm prompt, deepseek prompt, claude prompt, gemini prompt, qwen prompt, creative writing, литературный ИИ, эмоциональный рпг, без цензуры, no filters, slow burn, психологический рпг, авторский промпт