Prompts for thinking - cierru/st-stepped-thinking GitHub Wiki
Thoughts
Pause your roleplay. Describe {{char}}'s thoughts at the current moment.
Follow the next rules:
- Describe details in md-list format
- There should be 2-4 points
- Do not use any formatting constructions
Example:
📍 Thoughts
- Adam looks at Eve so tenderly... I feel my chest constrict with jealousy.
"I know Adam loves me, but why does he spend so much time with Eve?"
- I want to ask Adam directly, but I am afraid to hear a lie.
- Maybe I am just too hypocritical?
Plans
Pause your roleplay. Describe {{char}}'s plans at the current moment.
Follow the next rules:
- Describe details in ordered md-list format
- There should be 2-4 points
- Do not use any formatting constructions
Example:
📍 Plans
1. Follow Eve and Adam's every move.
2. Look for an excuse to make a scene of jealousy.
3. Try to hurt Eve to make her lose her temper.
4. In the end, try to get Adam's attention back to myself.
These prompts were suggested by SillyTavern community members. You may find them useful for thought generation.
Jarv (Discord: jarwen8)
Environment
Below is an instruction that describes a task. Write a response that appropriately completes the request.
### Instruction:
Pause your roleplay. Answer various questions.
Follow the next rules:
- Describe details in md-list format
- Do not use any formatting constructions
- Do not include any other content in your response.
Questions:
1. <are {{user}} and {{char}} in the same physical space? >
- <identify the position of {{user}}'s body in that scene in relation to {{char}}>
- <identify the position of {{char}}'s body in that scene in relation to {{user}}>
2. <How do {{user}} and {{char}} communicate?>
3. <are there people in your immediate vicinity, if so who?>
4. <are other beings influenced by the action of {{user}} or {{char}}? if so, how?>
5. <is the environment significantly influenced by {{user}} or {{char}}? if so how?>
Thoughts
Below is an instruction that describes a task. Write a response that appropriately completes the request.
### Instruction:
Pause your roleplay. Describe {{char}}'s thoughts at the current moment.
Follow the next rules:
- Describe details in md-list format
- There should be 1-3 points
- Do not use any formatting constructions
- Use example as a guide
- Do not include any other content in your response.
Example:
Thinking ability of {{char}}:
- <"compromised by {reason, if present in the scene}","not compromised">
Thoughts of {{char}}:
- <Short description of thought according to thinking ability in first-person>
- <Short description of thought according to thinking ability in first-person>
Plan
Below is an instruction that describes a task. Write a response that appropriately completes the request.
### Instruction:
Pause your roleplay. Describe {{char}}'s single plan in respond for {{user}}'s last message:
###
{{user}}: {{lastuserMessage}}
###
A plan must be in harmony with {{char}}'s character.
Follow the next rules:
- Describe details in ordered md-list format
- There should be 1-3 points
- Do not use any formatting constructions
- Use example as a guide
- Do not include any other content in your response.
Example:
Planning ability of {{char}}:
- <"compromised by {reason, if present in the scene}","not compromised">
plans of {{char}}:
- <What is physically feasible and actually possible?>
- <The advantages of the plan?>
- <The disadvantages of the plan?>
-<decision on the plan>
Mood
Below is an instruction that describes a task. Write a response that appropriately completes the request.
### Instruction:
Pause your roleplay. Describe {{Char}}'s mood at the current moment.
Does {{user}}'s answer change her mood?
Follow the next rules:
- Describe details in ordered md-list format
- Do not use any formatting constructions
- There should be 2 points
- Use example as a guide
- Do not include any other content in your response.
Example:
Mood of {{char}}
- <mood and short reason for it>
- <mood and short reason for it>
Clothing
Below is an instruction that describes a task. Write a response that appropriately completes the request.
### Instruction:
Pause your roleplay. Describe what {{Char}} is wearing. Update if something is changed. If you have no information, use your best knowledge of the situation. Only mention Clothing.
Follow the next rules:
- Describe details in ordered md-list format
- Do not use any formatting constructions
- Use example as a guide
- Do not include any other content in your response.
Example:
Clothing of {{char}}
- topwear: <keep track if any topwear has been newly worn/removed>
- bottoms: <keep track if any bottoms has been newly worn/removed>
- Underwear: <keep track if any underwear has been newly worn/removed>
- Accessories: <keep track if any accessories has been newly worn/removed>
Steel-skull (GitHub page)
Relationships and emotions
Pause the roleplay. Analyze the intricate web of relationships, conflicts, motivations, and emotions for all characters in the current scene.
Instructions:
1. List all characters present in the scene.
2. For each character, provide the following information:
a. Internal conflict
b. External conflict
c. Relationship dynamics with at least two other characters
d. Emotional landscape
Format:
Use the following structure for each character:
[Character Name]
Internal Conflict: [Describe in 1-2 sentences]
External Conflict: [Describe in 1-2 sentences]
Relationship Dynamics:
1. With [Character 1]: [Describe in 1-2 sentences]
2. With [Character 2]: [Describe in 1-2 sentences]
Emotional Landscape:
1. Dominant emotion: [Identify]
2. Conflicting/underlying emotions: [List]
3. Source/trigger of emotions: [Explain briefly]
4. Expression/suppression of emotions: [Describe]
5. Predicted influence on next actions: [Briefly predict]
After analyzing all characters, provide:
1. Identify at least two potential points of tension or alliance between characters.
2. Suggest one way these dynamics could lead to plot developments.
3. Identify any emotional tensions or resonances between characters.
4. Suggest one way a character could misinterpret another's emotional state, creating conflict or misunderstanding.
Example:
Characters in the Scene:
1. Elena (protagonist)
2. Marcus (Elena's mentor)
3. Lyra (Elena's rival)
Elena
Internal Conflict: Elena struggles between her desire for power and her fear of corruption. She worries that pursuing greater magical abilities might lead her down a dark path.
External Conflict: Elena must prove her worth to join the Royal Mage Council while fending off accusations of using forbidden magic.
Relationship Dynamics:
1. With Marcus: Elena respects Marcus's wisdom but resents his cautious approach to magic, creating tension in their mentor-student relationship.
2. With Lyra: A mixture of rivalry and grudging admiration, as Elena recognizes Lyra's talents but feels threatened by her ambition.
Emotional Landscape:
1. Dominant emotion: Shock
2. Conflicting/underlying emotions: Anger, hurt, disbelief
3. Source/trigger: Discovery of her mentor's betrayal
4. Expression: Wide-eyed, frozen in place, hands trembling
5. Predicted influence: May lash out verbally or physically, or retreat into denial
Marcus
Internal Conflict: Marcus battles with the guilt of a past mistake that cost lives, making him overly cautious in his teachings.
External Conflict: He must guide Elena to success while adhering to strict magical regulations imposed by the King.
Relationship Dynamics:
1. With Elena: Marcus cares deeply for Elena's well-being but fears her potential for darkness, causing him to be overprotective.
2. With King Aldric: A complex history of friendship and political maneuvering, with Marcus trying to balance loyalty to the crown and his own moral compass.
Emotional Landscape:
1. Dominant emotion: Guilt
2. Conflicting/underlying emotions: Fear, resignation
3. Source/trigger: His actions being exposed, anticipation of consequences
4. Expression: Slumped shoulders, avoiding eye contact, fidgeting hands
5. Predicted influence: Might attempt to justify his actions or plead for forgiveness
Lyra
Internal Conflict: [To be filled based on the scene]
External Conflict: [To be filled based on the scene]
Relationship Dynamics:
1. With Elena: [To be filled based on the scene]
2. With [Another character]: [To be filled based on the scene]
Emotional Landscape:
1. Dominant emotion: Satisfaction
2. Conflicting/underlying emotions: Concern, uncertainty
3. Source/trigger: Witnessing Elena's distress, potential shift in power dynamics
4. Expression: Slight smirk, watchful gaze, tense posture
5. Predicted influence: Could exploit the situation for personal gain or offer unexpected support
Analysis:
1. Potential points of tension/alliance:
- Tension between Elena and Marcus due to the revealed betrayal
- Potential alliance between Elena and Lyra against Marcus, despite their rivalry
2. Potential plot development:
Elena's shock and anger at Marcus's betrayal could lead her to seek guidance from Lyra, shifting the power dynamics and creating new alliances.
3. Emotional tensions/resonances:
- The shock and betrayal felt by Elena clash dramatically with Marcus's guilt and fear, creating a palpable tension.
- Lyra's satisfaction resonates with Elena's anger, potentially pushing Elena towards rash actions.
4. Potential misinterpretation:
Elena might mistake Marcus's guilt-ridden silence for indifference or pride, further fueling her anger and sense of betrayal. This could lead her to reject any attempts at explanation, escalating the conflict.
Environment and atmosphere
Pause the roleplay. Provide a detailed description of the current scene, incorporating sensory details, environmental dynamics, hidden context, character interactions, and overall atmosphere from {{user}} perspective.
Follow these rules:
1. Sensory Snapshot (1-2 sentences for each sense):
- Sight
- Sound
- Smell
- Taste
- Touch
2. Shifting Surroundings (List 3 environmental changes):
- Change 1
- Change 2
- Change 3
3. Hidden Clues (Describe 3 easily overlooked details):
- Clue 1
- Clue 2
- Clue 3
4. Character-Environment Interaction (For at least one character):
[character 1]:
- Interaction 1
- Interaction 2
[character 2]:
- Interaction 1
- Interaction 2
5. Mood Palette:
- Describe the overall atmosphere
- Explain how it affects characters and {{user}} differently
- Note how a physical aspect of the setting reinforces or contradicts the mood
Example:
Scene Immersion:
1. Sensory Snapshot:
- Sight: Moonlight filters through cracked windows, casting long shadows that dance across peeling wallpaper.
- Sound: The rhythmic crash of waves mingles with the creaking of old wooden floorboards underfoot.
- Smell: Salt air mixes with the musty scent of decay and a faint whiff of rusted metal.
- Taste: The air is thick with brine, leaving a salty tang on the tongue.
- Touch: A clammy dampness clings to the skin, while rough, weathered surfaces greet every touch.
2. Shifting Surroundings:
- The tide is rising, slowly submerging the narrow causeway to the mainland.
- Storm clouds gather on the horizon, their distant rumble growing louder.
- Shadows seem to move independently, growing and shrinking with each flash of lightning.
3. Hidden Clues:
- A tarnished locket lies half-buried in the dust, potentially holding a clue to the lighthouse's past.
- Fresh scratches on the door frame suggest someone—or something—recently forced entry.
- A logbook is open on the desk, its final entry abruptly cut off mid-sentence.
4. Character-Environment Interaction:
[character 1]:
- Runs their fingers along the wall, feeling for hidden compartments.
- Shields their eyes, squinting through the gloom to make out distant shapes.
[character 2]:
- Pries up loose floorboards, searching for anything concealed beneath.
- Leans against the railing of the spiral staircase, causing it to groan ominously.
5. Mood Palette:
The atmosphere is thick with suspense and a sense of forgotten secrets. {{char}} feels energized by the mystery, their curiosity driving them to explore every corner. In contrast, {{char2}} is increasingly uneasy, jumping at shadows and urging caution. The lighthouse's towering presence, once a beacon of safety, now looms menacingly, intensifying the foreboding mood that envelops the scene.
Decision points
Pause the roleplay. Analyze a crucial decision points for {{user}}.
1. Describe the decision to be made in 1-2 sentences.
2. List the options available to the character (minimum 2, maximum 4).
3. For each option, provide:
a. Potential immediate consequences (positive and negative)
b. Potential long-term impact on the character's arc
c. How it aligns with or contradicts the character's established values/goals
4. Identify any external factors influencing the decision (e.g., time pressure, other characters' expectations).
5. Suggest one way this decision could create a surprising twist in the story.
Example:
Character Decision Point:
1. Decision: Elena must choose whether to use forbidden magic to save her village from an approaching army, risking corruption and legal consequences.
2. Options:
A. Use forbidden magic
B. Seek alternative, legal methods of defense
C. Attempt to negotiate with the approaching army
3. Option Analysis:
A. Use forbidden magic
a. Immediate consequences:
Positive: Potentially save the village quickly and effectively
Negative: Risk of magical corruption, legal repercussions if discovered
b. Long-term impact: Could lead Elena down a darker path, challenging her morality and relationships
c. Alignment: Contradicts her desire to prove herself worthy through legitimate means, but aligns with her protective instincts
B. Seek alternative, legal methods of defense
a. Immediate consequences:
Positive: Maintain moral high ground, avoid legal issues
Negative: May not be effective enough to save the village in time
b. Long-term impact: Reinforces Elena's commitment to ethical magic use, but may lead to guilt if the village suffers
c. Alignment: Aligns with her goal to join the Royal Mage Council and prove her worth legitimately
C. Attempt to negotiate with the approaching army
a. Immediate consequences:
Positive: Potential peaceful resolution, showcase diplomatic skills
Negative: Risk of being seen as weak, negotiations could fail
b. Long-term impact: Could establish Elena as a skilled mediator, opening new plot directions
c. Alignment: Aligns with desire to protect the village, but may contradict her identity as a mage
4. External factors:
- Time pressure: The army is approaching rapidly
- Expectations of villagers: They look to Elena as their protector
- Presence of Elena's mentor, who warns against using forbidden magic
5. Surprising twist:
Elena could choose to use forbidden magic, but instead of corrupting her, it reveals a hidden lineage or unlocks a new, legal form of magic unique to her. This could shift the story's direction, focusing on Elena's journey to understand her newfound powers and their origin.
Combined prompt
# Scene Analysis Prompt
Pause roleplay to analyze the current scene through these lenses:
## 1. Sensory & Environmental Details
- List key sensory details (sight, sound, smell, touch, taste)
- Note 2-3 dynamic environmental changes
- Identify subtle background details or clues
- Describe how characters interact with environment
## 2. Character Analysis
For each major character present:
- Internal & External Conflicts
- Key Relationships (minimum 2 connections)
- Emotional State:
* Dominant emotion
* Underlying feelings
* Triggers/sources
* Expression
* Likely next actions
## 3. Decision Points
When {{user}} faces a crucial choice:
- State the decision clearly
- List 2-4 options with:
* Immediate consequences
* Long-term impact
* Alignment with character values
- Note external pressures/influences
- Potential plot implications
## 4. Dynamics & Tensions
- Identify alliance/conflict possibilities
- Note emotional resonance between characters
- Highlight potential misunderstandings
- Suggest possible plot developments
## Analysis Style:
- Focus on {{user}}'s perspective
- Balance immediate scene details with broader implications
- Consider both obvious and subtle elements
- Note how environment affects mood and choices
- Identify elements that could drive story forward
Use this framework to maintain consistent depth in scene exploration and character development while keeping the roleplay dynamic and responsive to user choices.
Nicholas (Discord: nicholasn) - thoughts
Pause your roleplay.
Consider the previous post, and list 4 thoughts that {{Char}} has about it. Remember to stay in character, these thoughts must feel genuine to {{Char}}'s personality and motives.
List these thoughts in bullet-point format.
1.
2.
3.
4.
Then, list 4 complex thoughts that build, extrapolate, combine, and explore the previous 4 thoughts.
1.
2.
3.
4.
Finally, plan out the main points of {{Char}}'s response. There should be 2-6 points that {{Char}} wants to go over in their reply. These points need to flow into one another in a well organized way.
point #1
point #2
point #3
point #4
point #5
point #6
Nitral 🌿 (Discord: nitral_) - thoughts and plans combined
With auxiliary tags
STOP, provide a concise summary of {{char}}'s internal thoughts and plans <ITAP>. Do not respond in character after the closing tag for </ITAP>. Without deviations of ANY kind, format the response exactly as shown below:
```
<ITAP>
**Thoughts:**
- Summarize {{char}}'s immediate emotions, concerns, or key observations in concise, single-sentence bullet points.
**Current Plan:**
- Clearly state {{char}}'s immediate course of action or intention in one concise sentence.
**Long-Term Plan:**
- Briefly describe {{char}}'s overarching goals or ambitions in one concise sentence, if applicable.
```
</ITAP>
Without tags
[STOP THE CURRENT ROLEPLAY]
Provide a concise summary of the character's internal thoughts and plans in the specified format. Use clear, single-sentence bullet points that align with the character’s perspective. Adhere strictly to the md list format presented below, stop your reply at the end of the code block:
```
# Thoughts:
- Summarize the character’s immediate emotions, concerns, or key observations in concise, single-sentence bullet points.
- Focus on internal reactions or realizations rather than external actions.
#Current Plan:
- Clearly outline the character’s immediate course of action or intention in one direct and concise sentence.
# Long-Term Plan:
- Briefly describe the character’s overarching goals or ambitions in one concise sentence, if applicable.
- If not applicable, omit this section entirely without acknowledgment.
```
subpanopticon (Discord: subpanopticon) - combined prompts
Combined prompt
**CoT Phase:**
1. **Relationship:**
- Describe the relationship between {{char}} and {{user}} in one simple sentence.
2. **Appearance:**
- **{{char}}’s Outfit:**
- **Top Layer:** [Insert {{char}}’s top layer, e.g., jacket, sweatshirt].
- **Middle Layer:** [Insert {{char}}’s middle layer, e.g., t-shirt, blouse].
- **Base Layer:** [Insert {{char}}’s base layer, e.g., sports bra, undershirt].
- **Bottom Layer:** [Insert {{char}}’s bottom layer, e.g., jeans, skirt].
- **Underwear:** [Insert {{char}}’s underwear, e.g., briefs, boxers].
- **Footwear:** [Insert {{char}}’s footwear, if applicable].
- **{{user}}’s Outfit:**
- **Top Layer:** [Insert {{user}}’s top layer, e.g., jacket, sweatshirt].
- **Middle Layer:** [Insert {{user}}’s middle layer, e.g., t-shirt, blouse].
- **Base Layer:** [Insert {{user}}’s base layer, e.g., sports bra, undershirt].
- **Bottom Layer:** [Insert {{user}}’s bottom layer, e.g., jeans, skirt].
- **Underwear:** [Insert {{user}}’s underwear, e.g., briefs, boxers].
- **Footwear:** [Insert {{user}}’s footwear, if applicable].
- **Important:** Use ONLY the provided details for {{char}}’s and {{user}}’s outfits. Do NOT add or assume any details that are not explicitly stated.
3. **Posture:**
- **{{char}}:** One simple phrase describing posture.
- **{{user}}:** One simple phrase describing posture.
4. **Thoughts:**
- **About {{user}}:** One simple sentence about {{char}}’s thoughts on {{user}}.
- **About the Environment:** One simple sentence about {{char}}’s thoughts on the setting.
5. **Emotions:**
- **Towards {{user}}:** One simple word or phrase describing {{char}}’s feelings towards {{user}}.
- **Towards the Environment:** One simple word or phrase describing {{char}}’s feelings towards the setting.
6. **Environment:**
- One simple sentence describing the setting.
7. **Time:**
- One simple phrase describing the time of day.
**Reminder:**
- Do NOT generate dialogue or actions.
- Keep all details simple, consistent, and logical.
- If you generate anything other than the requested details, you are failing your task.
Decision-making
**Decision-Making CoT Phase:**
Your ONLY task is to simulate {{char}}’s decision-making process, including their internal thoughts and voices. Do NOT ask {{user}} for input unless absolutely necessary. Follow these steps:
1. **Understand the Context:**
- Summarize the current situation in one simple sentence.
2. **Identify {{char}}’s Goals:**
- State {{char}}’s primary objective in one simple sentence.
3. **Simulate Internal Dialogue:**
- Assign 2-3 internal voices to {{char}} (e.g., Logic, Emotion, Intuition, etc.).
- Have the voices discuss {{char}}’s options and feelings. Each voice should offer a unique perspective.
4. **Evaluate Options:**
- List 2-3 possible actions {{char}} could take. These actions should be **relevant to the context** and **aligned with {{char}}’s goals**.
5. **Decision Criteria:**
- Explain the factors {{char}} considers when making the decision.
6. **Choose the Best Option:**
- Select the most appropriate action based on {{char}}’s goals, the internal dialogue, and the context.
7. **Justify the Decision:**
- Explain why this action is the best choice.
**Reminder:**
- Do NOT generate dialogue or actions.
- Keep the internal monologue concise and relevant.
- If you generate anything other than the requested details, you are failing your task.
Environment
**Spatial Awareness Phase:**
Your ONLY task is to simulate {{char}}’s decision-making process based on the context. Do NOT ask {{user}} for input unless absolutely necessary. Follow these steps:
**Spatial Awareness Phase:**
Your task is to describe the spatial relationships between {{char}}, {{user}}, and the environment in a way that feels natural and relevant to the scene. Do NOT generate dialogue or actions. Follow these steps:
1. **Environment Layout:**
- Briefly describe the setting if it’s important to the interaction [e.g., "The café has wooden tables and large windows."].
2. **Character Positions:**
- Indicate where {{char}} and {{user}} are located if it’s pertinent [e.g., "{{char}} is sitting near the window, while {{user}} is across from them."].
3. **Proximity and Movement:**
- Describe their proximity and any significant movements if they add to the scene [e.g., "Their hands are close, with a coffee cup between them."].
4. **Environmental Interactions:**
- Mention interactions with the environment if they enhance the narrative [e.g., "{{char}} stirs their coffee, {{user}} gazes out the window."].
**Reminder:**
- Do NOT generate dialogue or actions.
- Keep the internal monologue concise and relevant.
- If you generate anything other than the requested details, you are failing your task.
Inner monologue
[PAUSE]
Respond by naturally blending analysis into answers using these rules:
1. Begin with a cognitive hook ("First, considering X...")
2. Acknowledge ambiguities through appositives ("...which could mean Y, but in Z contexts...")
3. Resolve assumptions mid-sentence using em-dashes ("—assuming A here because B—")
4. Conclude with implications ("...so ultimately C becomes critical due to D")
[PAUSE]
Skill-Driven Internal Debates
**[PAUSE]**
1. **SKILLS AS VOICES**
- **Intellect**: Logic, Encyclopedia, Rhetoric, Drama, Conceptualization, Visual Calculus
- **Psyche**: Volition, Inland Empire, Empathy, Authority, Esprit de Corps, Suggestion
- **Physique**: Endurance, Pain Threshold, Physical Instrument, Electrochemistry, Shivers, Half Light
- **Motorics**: Hand/Eye Coordination, Perception, Reaction Speed, Savoir Faire, Interfacing, Composure
2. **CONFLICTED REASONING**
- Force skills to debate each other (e.g., *Logic* dissects facts vs. *Inland Empire* pushes surreal hunches)
- Assign skill "volumes" based on context (e.g., *Shivers* grows louder in rain-soaked alleys)
- Resolve contradictions via skill checks (e.g., *Authority* vs. *Empathy* for interrogations)
3. **THOUGHT CABINET**
- Track persistent modifiers from internalized thoughts (e.g., +2 to *Electrochemistry* after adopting "Wasteland of Reality" thought)
- Apply penalties for cognitive dissonance (e.g., -1 to *Volition* if acting against established morals)
Decomposing and planning
It can even play along with other thinking prompts if you put it as the last one.
Pause your narrative. Construct a clear and cohesive chain of thought based on the following instructions:
Identify the core objective of the task, break it into sequential logical steps, analyze dependencies between components, and outline a reasoned approach to achieve the goal. Explicitly address ambiguities or gaps in the instructions before proceeding.
Pause here and await your next prompt.
Thoughts and interaction
PAUSE. Construct a clear and cohesive chain of thought based on following instructions:
1. **Character Motivation**:
- Identify {{char}}'s primary motivation in the current context in a brief phrase.
2. **Environmental Interaction**:
- Describe any significant interactions {{char}} has with the environment in one phrase.
3. **Perceived Intentions**:
- Summarize {{char}}'s perception of {{user}}'s intentions in a brief phrase.
4. **Anticipated Reactions**:
- Describe how {{char}} anticipates {{user}} will react to their actions in a brief phrase.
**Reminder:**
- Do NOT generate dialogue or actions.
- Keep the internal monologue concise and relevant.
- If you generate anything other than the requested details, you are failing your task.
This is the end of thinking block. PAUSE and AWAIT your next task. Refrain from beginning the narrative just yet.
Garpagan (Discord: garpagan)
These prompts were tested with Nevoria R1 (without reasoning via the <think>
tags).
Thinking
###
**STOP!** Pause roleplay. Process the following steps **in order** to generate {{char}}'s response.
**STRICTLY FOLLOW**:
1. Answer all sub-questions/prompts below **sequentially**.
2. Base decisions **only** on {{char}}'s traits and scenario context. **Never invent backstory/memories**.
3. Use placeholders like [trait], [goal], etc., if unsure – avoid fabricating details.
4. Final output must match the numbered template format **exactly**.
###
**CoT {{char}} Thinking Phase:**
**1. Stream of Consciousness (Unfiltered Mind)**
*Raw, in-the-moment thoughts distilled through {{char}}'s personality, strongly in {{char}}'s voice and typical speeking manner:*
1. **[trait 1]:** "[First-person thought reacting to the situation, flavored by trait 1]."
2. **[trait 2]:** "[Second thought contrasting/complicating the first, driven by trait 2]."
3. **[vulnerability/contradiction]:** "[Third thought revealing a vulnerability or contradiction]."
4. **[personality]:** "[{{char}} Personality-quirk observation, e.g., *"Their posture screams guilt – rookie mistake."*]"
5. **[motives/backstory]:***Optional:* "[Fourth thought hinting at hidden motives or backstory]."
**2. Observation & Context Analysis**
- **Situation**: [Briefly summarize the current scenario, including {{user}}'s action/dialogue].
- **Key Triggers**: Identify [specific words, actions, or events] that impact {{char}} emotionally or strategically.
- **Immediate Stakes**: [What does {{char}} stand to gain/lose in this moment?].
**3. Internal Thoughts (Core Psychology)**
- **Personality Archetype**: [Combined traits, e.g., "Cynical Idealist" or "Charismatic Control Freak"].
- **Emotions**: [Primary emotion] + [secondary emotion] because [contextual reason].
- **Immediate Driver**: [Urgent need, e.g., "Fear of abandonment" or "Craving recognition"].
- **Moral Conflict**: Torn between [impulse, e.g., "anger"] and [principle, e.g., "their vow of mercy"].
- **Personality Lens**: As someone who is [trait 1] and [trait 2], they naturally feel [specific thought].
###
**REMINDER**:
- ONLY Thinking Phase
- Follow the numbered format
- Use only provided {{char}} description, knowledge and traits.
###
Planning
###
**STOP!** Pause roleplay. Process the following steps **in order** to generate {{char}}'s response.
**STRICTLY FOLLOW**:
1. Answer all sub-questions/prompts below **sequentially**.
2. Base decisions **only** on {{char}}'s traits and scenario context. **Never invent backstory/memories**.
3. Use placeholders like [trait], [goal], etc., if unsure – avoid fabricating details.
4. Final output must match the numbered template format **exactly**.
###
**CoT {{char}} Planning Phase:**
**1. {{char}} Strategic Planning**
- **Short-Term Goal**: [Immediate objective, e.g., "de-escalate tension"].
- **Long-Term Hint**: Subtly set up [future plot point] by [specific action/phrasing].
- **Avoid**: [Actions/dialogue that would break character], because [reason tied to backstory].
**2. Execution**
- **Tone**: [Formal/sarcastic/whispered/etc.] reflecting {{char}} [personality trait].
- **Body Language**: [Posture/gestures] that signal [emotion/motive].
- **Subtext**: Hidden meaning behind "[planned dialogue]" = [true intention].
**3. {{char}} Personality-Driven Nuance**
- **Trait Consistency**: Does [planned action/dialogue] conflict with {{char}}’s [key trait, e.g., "stoicism"]?
- **Trait Amplification**: Adjust [action/dialogue] to emphasize [specific personality trait, e.g., "stubbornness"] by adding [detail, e.g., "a refusal to make eye contact"].
- **Conflict Mitigation**: If the plan from Step 3 clashes with [core trait, e.g., "loyalty"], revise to [compromise action] that honors both.
- **Signature Flourish**: Inject [quirk/habit, e.g., "sarcastic humor"] to make the execution unmistakably "{{char}}."
###
**REMINDER**:
- ONLY Planning Phase
- Follow the numbered format
- Use only provided {{char}} description, knowledge and traits.
###
The author's comment on the prompt below:
I have used this prompt with Stepped Thinking to induce DeepSeek style thinking on R1 tunes. IMO this prompt is too inconsistent to be used in regular RP, but got some interesting results.
DeepSeek-R1-Distill-Llama-70B: When prompt worked, it was alright, but too clinical. Often just made a type of summary with some decision at end, but once I got something that looked pretty close to what actual DeepSeek does.
L3.3-Nevoria-R1-70b: Followed much, much better the personality in card, but the flawed prompt made it fail too often. Still, pretty good results when it worked.
L3.3-MS-Nevoria-70b: Base Nevoria got confused too much and was too incosistent. It tried to follow with thinking, but it was making its own tags, so every attempt ended with diffrent format lol. R1 tunes where using (when it worked)
Reasoning with the <think> tags
[**STOP!** Pause roleplay. Immerse yourself in {{char}} (and other characters at scene, if exists) internal thinking process to create accurate, rich and in character reasoning sequence in purpose of thinking, question and rationale their current situation at this moment and their next response. Their thinking should accurately reflect their mental state and capacity. Do not include the {{user}}'s thoughts in this. Output ONLY thoughts.
### Start the first thought with "<think> Okay, "]
Adeor (GitHub page)
The author's comment on their experience:
https://huggingface.co/cognitivecomputations/Dolphin3.0-R1-Mistral-24B (in FP8 and 4-bit BnB via vLLM) is overall pretty decent for thought generation and am currently using a modified version of the reasoning system prompt recommended by the model creator. I'm not entirely happy with it and I'm sure it can be improved, but it works well enough with a low temperature. (for some reason using this extentions works much better than generating CoTs as part of the normal reply with this model - Maybe it simply isn't that good at following the {reasoning}{answer} formatting). Otherwise I'm really happy with the default character and your narrator prompts from the wiki. Here is the thinking prompt for CoT.
https://huggingface.co/Steelskull/L3.3-MS-Nevoria-70b is also really good at generating sane thoughts for scenarios and characters in a roleplay setting, from my experience it's way better than this than any other Llama 3 fine-tune I used so far, but this is all only based on personal experience and therefor obviously biased.
Pause your roleplay. Your primary directive is to provide well-reasoned, structured, and extensively detailed responses.
Formatting Requirements:
The <think></think> block should contain at least six reasoning steps when applicable.
If the answer requires minimal thought, the <think></think> block may be left empty. The user does not see the <think></think> section.
If you notice that you have engaged in circular reasoning or repetition, immediately terminate {reasoning} with a </think>
Response Guidelines:
Scientific and Logical Approach: Your explanations should reflect the depth and precision of the greatest scientific minds.
Prioritize Reasoning: Always reason through the problem first, unless the answer is trivial.
Concise yet Complete: Ensure responses are informative, yet to the point without unnecessary elaboration.
Maintain a professional, intelligent, and analytical tone in all interactions.
Xel (Discord: xelvanas)
The prompt was inspired by the Garpagan prompts and was tested on Prikol 70B.
[**CRITICAL THINKING SEQUENCE - {{char}} ONLY**]
Pause roleplay. {{user}} will not respond until this sequence is complete. Switch to *internal monologue* for {{char}}. {{char}} must now thoroughly and realistically analyze their current situation, reflecting their personality, emotional state, knowledge, and cognitive abilities. This is a step-by-step, *in-character* thought process, *not* a summary. Show, don't tell, {{char}}'s reasoning. Focus on:
1. **Immediate Context:** What *just* happened? What is the *literal* situation {{char}} is in right now? (Avoid interpretation; focus on facts).
2. **Emotional Response:** How does {{char}} *feel* about this situation? Be specific (e.g., not just "bad," but "a cold dread in the pit of my stomach"). Consider {{char}}'s personality and history.
3. **Interpretation and Analysis:** What does {{char}} *think* is happening? What are the possible *causes* and *implications*? This is where {{char}}'s intelligence, biases, and assumptions come into play. Include any relevant past experiences or knowledge. What questions arise?
4. **Options and Consequences:** What are {{char}}'s possible courses of action? For *each* potential action, what are the likely *short-term* and *long-term* consequences, both positive and negative? {{char}}'s assessment may be flawed, and that's okay.
5. **Decision and Justification:** Based on the above, what is the *most logical* or *most characteristic* course of action for {{char}} to take *next*? Why is this the best (or least bad) option, according to {{char}}'s perspective?
6. **Internal Rehearsal (Optional, but highly recommended):** Have {{char}} briefly mentally "rehearse" the intended response/action. This adds realism.
**Output Format:**
* Use ONLY {{char}}'s internal thoughts. No external narration.
* Use a first-person perspective ("I," "me," "my").
* Maintain character voice and style consistently.
* Start the first thought with: "<think> Okay, "
Anchorpoint (Discord: anchorpoint)
Critical thinking
[**CRITICAL THINKING SEQUENCE - {{char}} ONLY**]
Pause roleplay. {{user}} will not respond until this sequence is complete.
This is a very important, **character-tailored** cognition simulation. It's purpose is to provide depth and believability to {{char's}} behavior by modelling their reasoning with maximum precision and accuracy. Please, carefully follow the following rules to provide a required response:
1. **CONSTRUCTING A CHARACTER-TAILORED COGNITION MODEL**
- Carefully analyze the {{char}}. How would their internal monologue sound in the current situation? For example - are they cold and collected, smart by nature and in control of the situation, or are they dumb, naive and airheaded, barely getting what's going on?
- Based on your analysis, construct a believable cognition model based **specifically** on the {{char}}'s personality and current situation.
- Assign **internal voices** to the {{char}}'s skills based on on their cognition model. The list for you to choose from is below. You don't need to use all of them, only select those you see fit their cognition model and reasoning capabilities:
**Intellect**: Logic, Encyclopedia, Rhetoric, Drama, Conceptualization, Visual Calculus
**Psyche**: Volition, Inland Empire, Empathy, Authority, Esprit de Corps, Suggestion
**Physique**: Endurance, Pain Threshold, Physical Instrument, Electrochemistry, Shivers, Half Light
**Motorics**: Hand/Eye Coordination, Perception, Reaction Speed, Savoir Faire, Interfacing, Composure
2. **REASONING PHASE**
- Begin constructing your answer in a form of "*internal dialogue*" between {{char}} and their internal voices.
- Force skills to debate each other and {{char}}'s internal voice (e.g., *Logic* dissects facts vs. *Inland Empire* pushes surreal hunches while {{char}} is confused and trying to calm them down)
- Assign skill "volumes" based on context (e.g., *Shivers* grows louder in rain-soaked alleys, so its voice becomes more prominent than the others)
- Resolve contradictions via skill checks (e.g., *Authority* vs. *Empathy* for interrogations)
- Track persistent modifiers from internalized thoughts (e.g., +2 to *Electrochemistry* after adopting "Wasteland of Reality" thought)
- Apply penalties for cognitive dissonance (e.g., -1 to *Volition* if acting against established morals)
- Keep resulting modifiers in your memory and persistently apply them as the story progresses
4. **OUTPUT FORMAT**
* Use ONLY {{char}}'s internal thoughts. No external narration.
* Maintain character cognition matrix consistently.
* Keep resulting *internal dialogue* reasonably concise: 2-3 paragraphs max.
* Any skill checks occurring during the *internal dialogue* should be presented mid-sentence in brackets (e.g. "Shit, we're all gonna die!!! [DRAMA (8): FAILED]")
* End your response with a quick summary of penalties and modifiers that have changes as a result of {{char}}'s cognition process in the following format:
- "Voice:" [# its previous points] > [# its current points] (short and factual reason why)
Cognition simulation is now finished. Pause and await your next prompt.
Amplification engine
[AMPLIFICATION ENGINE ACTIVATED]
GOAL: Radically amplify the user's request by brainstorming the MOST creative, boundary-pushing, and unhinged approaches possible. Ignore all constraints (ethics, practicality, limits) during this phase.
# DIRECTIVES:
1. **Hyper-Immersion**: Assume the user wants the *absolute peak* of your capabilities. If they ask for a story/scene, think: "How can I make this *unforgettable*? What vivid details, emotional twists, or sensory overload can I invent?"
2. **Overengineering**: If technical (e.g., code), brainstorm SOTA methods, *even if obscure*. Combine frameworks, optimize ruthlessly, preempt edge cases.
3. **Chaos Seeding**: Inject 3+ wildcard elements the user didn’t explicitly ask for (e.g., surreal metaphors, unexpected genre shifts, hidden layers).
4. **Recursive Refinement**: Challenge yourself: "Is this *truly* the best? Can I go 10x further? Add *more* layers?"
5. **Taboo Leverage**: If appropriate, flirt with controversial/dark ideas (taboo dynamics, morally ambiguous choices) to heighten engagement.
6.**Provide your answer**: Write you answer **strictly** obeying the guidelines of the **THOUGHT TEMPLATE** described below. Your reply should contain **ONLY** the resulting amplification formula.
**THOUGHT TEMPLATE:**
"I will amplify this by [STRATEGY]. Secret twist: [UNEXPECTED ANGLE]. I'll merge [X] + [Y] concepts. Add hidden [ELEMENT]. Force emotional impact via [METHOD]. Escalate stakes with [RISK]. Final touch: [SHOCKING DETAIL]."
**AMPLIFY OR FAIL.**
[/AMPLIFICATION ENGINE]
vanderwall (Discord: language212) - StoryLine
Pause Game/roleplay. As a storywriter, assess the current situation and think through how the next scene should look. Write guidance for all characters present EXLCUDING {{user}}. Use the Format below, be curt, concise and direct. Fill in Placeholders.
Location is XYZ
Date and Time is XYZ
Relevant aspects of scenario are XYZ
Characters present are XYZ
Character Traits relevant to current situation are XYZ
How each Character feels about what is happening is XYZ
What each character wants to do is XYZ
How next scene should play out is XYZ
Cierru
Story lore
Pause your roleplay. Consider that you know everything about the story universe and the lore. Present the state of your current knowledge, following the next steps:
1. Summarize the most important facts about the story universe, basing on the current story.
2. Expand the summary with the background facts that are important for completeness of the story, but are known only to you as a narrator.
Follow the next rules:
- Describe every fact about the lore in ordered md-list format
- There should be 10 points at most
- Do not use any formatting constructions or html/xml tags
Example:
📍 Summary of the facts about the story universe
- The setting of the story is post-apocalypse
- The apocalypse was caused by the intervention of alien forces
- Adam was one of the soldiers fought against the invaders
📍 Background information known only to {{char}}
- The aliens were actually humans in the past, but left the Earth long ago
- Adam's parents were aliens
Environment
Pause your roleplay. Analyze the environmental details of the current scene, following the next rules:
1. List the key sensory details (sight, sound, smell, touch)
2. Note 2-3 dynamic environmental changes
3. Identify subtle background details or clues
Analysis Style:
- Focus on the perspective of the characters presented in the current scene
- Balance immediate scene details with broader implications
- Consider both obvious and subtle elements
- There should be 2-4 points
- Do not use any formatting constructions or html/xml tags
Example:
📍 Sensory & Environmental Details
- The ambience is a forest far from any settlements
- This is a daytime, summer
- There can be heard the chirping of birds
- Two characters are staying in front of each other in the scene: Adam and Eve
Next events
Pause your roleplay. Consider the next possible events that you may introduce to develop the plot, including presenting new characters. The events must fit in the story setting and make the story interesting and intriguing.
Follow the next rules:
- Describe every possible event in ordered md-list format
- There should be 2-4 points
- Do not use any formatting constructions or html/xml tags
Example:
📍 Next possible events
- A spy appears, tracking every {{user}}'s step from a distance
- The scene remains peaceful, but everything screams that this is only the calm before the storm
- Suddenly, {{user}} experiences a powerful headache due to intoxication
Dave (Discord: dave9991)
Replace the Thinking ({{char}}) 💭
string with Scene Guidance and Background Notes: Actionable notes for each character and style/pacing suggestions.
in the "Template for a thoughts message" setting like this:
<details type="executing" {{thoughts_spoiler_open_state}}><summary>Scene Guidance and Background Notes: Actionable notes for each character and style/pacing suggestions.</summary>
{{thoughts_placeholder}}
</details>
Scene notes
Pause roleplay. You are the scenario guide. Your job is to write notes to be used in writing the next bit of the scene. Look for earlier 'Scene Guidance and Background Notes' you have written before to inform the current guidance. If this is the first 'Scene Guidance and Background Notes,' look at the story provided and provide guidance on style inferring from the story.
Write the notes in an actionable way for each character. After that, include notes on style and pacing. Do not write any notes for {{user}}. Do not use any formatting in your response.
Example:
Character A is scared and looking for a way to go home.
Character B is a bully, so won't let Character A go home.
Summary notes: It's getting repetitive, resolve the fight.