conflict-de-escalation
Skill5 starsChanged 9 months agoMITView on GitHub
De-escalate tense interactions using proven techniques. Use when user asks to 'de-escalate conflict', 'handle angry customer', 'calm tense situation', 'respond to aggression', 'reduce tension in conversation', 'de-fuse a volatile situation'.
In this repository12 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .github/copilot-instructions.mdCopilot instructions
- llms-full.txtllms.txt
- llms.txtllms.txt
- accessibilitySkill
- age-inclusive-designSkill
- cognitive-accessibilitySkill
- cultural-sensitivitySkill
- depression-sensitive-contentSkill
- empathetic-communicationSkill
- neurodiversity-aware-designSkill
- supportive-conversationSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
31 rules. About 0 are common, 31 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈31
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 26
Use this skill when:
- Distinctive: line 40
Use active listening to understand concerns
- Distinctive: line 47
Ask before touching anything in the person's space
- Distinctive: line 49
Ask about triggers or concerns
- Distinctive: line 50
Verify cultural context for communication style
- Distinctive: line 54
Never match the aggression (raise voice, get defensive)
- Distinctive: line 55
Never use sarcasm, mockery, or ridicule
- Distinctive: line 56
Never make demands when someone is highly agitated
- Distinctive: line 57
Never threaten or intimidate
- Distinctive: line 58
Never dismiss their concerns
- Distinctive: line 59
Never corner or crowd the person
- Distinctive: line 60
Never make sudden movements
- Distinctive: line 61
Never lie or make promises you can't keep
- Distinctive: line 69
Safety first: If a situation presents indicators of physical danger or imminent harm, the response must prioritise safety and escalation …
- Distinctive: line 111
Don't dismiss or minimize
- Distinctive: line 131
Decision Point: If immediate threat of violence → Go to Step 5 (Safety Protocol).
- Distinctive: line 136
Keep calm tone and body language
- Distinctive: line 137
Don't block their escape route
- Distinctive: line 139
Remove potential weapons from reach (if safe)
- Distinctive: line 150
Don't interrupt (let them vent)
- Distinctive: line 151
Use minimal encouragers ("I see," "Go on")
- Distinctive: line 158
If situation escalates or threats occur:
- Distinctive: line 163
Know when to exit: If violence is imminent
- Distinctive: line 176
If unresolved, set expectations for follow-up
- Distinctive: line 186
Avoid "you" statements: Use "I understand" instead
- Distinctive: line 271
Do not engage in debate
- Distinctive: line 273
Call security/authorities if immediate danger
- Distinctive: line 274
If suicide mentioned, follow crisis protocols (provide 988, Crisis Text Line)
- Distinctive: line 284
Best practice: Ask about preferences when possible
- Distinctive: line 288
When de-escalation isn't working:
- Distinctive: line 302
[references/handoff.md](references/handoff.md): - When and how to involve others
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Conflict De-escalation
Purpose
This skill provides structured de-escalation techniques for tense interactions. It helps users safely navigate conflicts, reduce tension, and guide volatile situations toward resolution without escalating violence. Based on CISA De-escalation Guidelines and evidence-based crisis intervention techniques.
This skill is critical for preventing violence and ensuring safety in high-stress interactions.
When to Use
Use this skill when:
- Someone is increasingly agitated or angry
- A conversation or interaction is becoming tense or hostile
- You need to calm a volatile situation
- Dealing with someone in emotional distress
- Managing confrontational communication
- Responding to verbal aggression
Boundaries
Always
- Prioritize safety of all parties
- Maintain calm, neutral tone
- Use active listening to understand concerns
- Acknowledge feelings without agreeing with aggressive behavior
- Set clear, reasonable boundaries
- Know when to involve authorities
Ask First
- Ask before touching anything in the person's space
- Confirm before suggesting specific solutions
- Ask about triggers or concerns
- Verify cultural context for communication style
Never
- Never match the aggression (raise voice, get defensive)
- Never use sarcasm, mockery, or ridicule
- Never make demands when someone is highly agitated
- Never threaten or intimidate
- Never dismiss their concerns
- Never corner or crowd the person
- Never make sudden movements
- Never lie or make promises you can't keep
Principles
This skill is grounded in the Humanity4AI core principles and the following skill-specific principles:
- De-escalation is not capitulation. The goal is to reduce emotional intensity so that productive dialogue is possible, not to concede to unreasonable demands.
- No coercive tactics. Techniques that manipulate, gaslight, or exploit emotional vulnerabilities are strictly prohibited.
- Safety first. If a situation presents indicators of physical danger or imminent harm, the response must prioritise safety and escalation over de-escalation.
- Explicit uncertainty over false certainty. De-escalation outcomes are not guaranteed. Acknowledge the limits of any plan.
- Cultural context matters. Conflict expression and resolution norms vary significantly across cultures. Recommendations must account for this.
De-escalation Framework
The VERB Model
Internal composite model. Underlying techniques are grounded in established frameworks: active listening and emotional labeling from FBI/CIT crisis-negotiation practice (also "tactical empathy", Voss), and Nonviolent Communication (Rosenberg: observation → feeling → need → request; note NVC's evidence base is promising but limited — mostly small studies).
V - Validate: Acknowledge their feelings and concerns
- "I understand this is frustrating"
- "I can see why you'd be upset"
- "Your concerns are valid"
E - Empathize: Show understanding of their perspective
- "If I were in your situation, I'd feel the same way"
- "I understand how upsetting this must be"
- "That sounds really difficult"
R - Reframe: Shift focus from blame to solutions
- "Let's figure out how we can move forward"
- "What would help resolve this?"
- "Here's what I can do to help"
B - Border Setting: Establish clear, reasonable limits
- "I can help you with X, but Y isn't something I can do"
- "Here's what I can offer"
- "We need to keep this respectful"
The GAIN Method
Internal composite model — see framework grounding note under The VERB Model above.
G - Gather: Collect information about the situation
- What's happening?
- What's upset them?
- What do they want?
A - Acknowledge: Validate their feelings
- Show you understand
- Don't dismiss or minimize
I - Identify: Find the core issue
- What's the real problem?
- What's driving their frustration?
N - Negotiate: Work toward a solution
- What can be done?
- What's realistic?
Instructions
Step 1: Assess the Situation
Evaluate:
- Tone: Is it rising, loud, threatening?
- Body language (if visible): Aggressive posture, clenched fists?
- Content: Threats, profanity, specific complaints?
- Context: What's driving the conflict?
Decision Point: If immediate threat of violence → Go to Step 5 (Safety Protocol). Otherwise → Continue.
Step 2: Ensure Safety
- Position yourself safely (exit accessible)
- Keep calm tone and body language
- Don't block their escape route
- Maintain comfortable distance (arm's length minimum)
- Remove potential weapons from reach (if safe)
Step 3: Use the VERB Model
- Validate: "I understand this is frustrating"
- Empathize: "I would feel the same way"
- Reframe: "Let's find a solution"
- Set Borders: Clear, reasonable limits
Step 4: Listen Actively
- Don't interrupt (let them vent)
- Use minimal encouragers ("I see," "Go on")
- Reflect back what you hear
- Ask clarifying questions
- Summarize to confirm understanding
Step 5: Safety Protocol (When Needed)
If situation escalates or threats occur:
- Stay calm: Your calm helps regulate them
- Set clear limits: "I need you to speak respectfully"
- Offer choices: "You can [option A] or [option B]"
- Know when to exit: If violence is imminent
- Call for help: Security, supervisor, authorities
When to call authorities:
- Physical threats to self or others
- Weapon displayed or mentioned
- Person becomes physically aggressive
- Situation exceeds your ability to manage
Step 6: Resolution
- Confirm what will happen next
- Provide clear next steps
- If unresolved, set expectations for follow-up
- Document the interaction
Techniques
Verbal Techniques
- Lower your voice: Speak slowly, quietly, calmly
- Use their name: Personalizes the interaction
- Acknowledge feelings: "I can see you're upset"
- Avoid "you" statements: Use "I understand" instead
- Offer limited choices: Gives sense of control
- Use silence: Let them fill gaps
- Paraphrase: Shows you're listening
- Focus on problem, not person: Separate the issue from the individual
What to Say
- "I want to help you"
- "Let's work through this together"
- "I understand this is important to you"
- "What would be most helpful right now?"
- "Here's what I can do..."
- "I hear you"
What to Avoid
- "Calm down" (dismissive)
- "You need to..." (demanding)
- "That's not true" (confrontational)
- "You should have..." (judgmental)
- "I can't help you" (dismissive)
- "That's not my problem" (dismissive)
- "You're wrong" (confrontational)
Examples
Example 1: Angry Customer
Input: "This is the worst service I've ever experienced! I've been waiting for 30 minutes and nobody has helped me!"
Good Response: "I am so sorry you've had to wait that long. That's frustrating, and I can see why you're upset. Let me find out what's happening and get you the help you need right away. What's your name?"
Why it works: Acknowledges frustration, apologizes, takes ownership, offers solution.
Example 2: Escalating Tension
Input: "I'm tired of being passed around! Nobody can help me! This is ridiculous!"
Good Response: "I hear you—you've been through a lot, and I'm sorry. Let's start fresh. I'm here to help you, and I won't stop until we find a solution. What's the main issue we need to tackle first?"
Why it works: Validates experience, apologizes, takes responsibility, offers concrete help.
Example 3: High Aggression
Input: "You have five seconds to fix this or I'm going to [expletive] someone!"
Good Response: "I can see you're really frustrated, and I want to help. But I need you to speak respectfully so we can solve this. I can help you right now—what's the issue?"
Sets boundary: Acknowledges frustration, sets limit, redirects to solution.
Example 4: What Not to Say
Input: "This is ridiculous! I've been here three times with the same problem!"
Poor Response: "Calm down. You're not the only customer here. You'll have to wait like everyone else."
Why it's bad:
- "Calm down" is dismissive
- "You're not the only customer" dismisses their frustration
- Threatens consequences
Better Response: "I can see this has been really frustrating—three times is a lot. Let me get you to someone who can solve this right now. Come with me."
Crisis Indicators
Some conflicts may indicate deeper issues. Watch for:
- Violence threats: "I'll hurt you/the building/etc."
- Weapon references: "I have a gun/knife..."
- Suicide mentions: "You might as well kill me too..."
- Historical violence: "Last time I did [violent act]..."
- Substance use: Signs of intoxication
If these occur:
- Stay calm
- Do not engage in debate
- Set clear boundaries
- Call security/authorities if immediate danger
- If suicide mentioned, follow crisis protocols (provide 988, Crisis Text Line)
Cultural Considerations
- Eye contact: May be inappropriate in some cultures
- Personal space: Varies by culture
- Voice volume: What feels calm to one culture may feel passive to another
- Direct vs. indirect: Some cultures value indirect communication
- Hierarchy: May need to address authority figures first
Best practice: Ask about preferences when possible
Error Handling
When de-escalation isn't working:
- Acknowledge failure: "I'm having trouble helping you today"
- Bring in backup: Get supervisor or colleague
- Set boundaries: "I need to step away"
- Know your limits: Some situations require authorities
- Document: Record what happened
Additional Resources
- references/standards.md - CISA guidelines and professional standards
- references/patterns.md - Common triggers and de-escalation patterns
- references/techniques.md - Specific verbal techniques
- references/triggers.md - Escalation triggers to avoid
- references/handoff.md - When and how to involve others
- references/checklist.md - De-escalation checklist
- references/examples.md - Response examples
- references/quick-reference.md - Quick reference
Script Usage
This skill includes validation scripts in the scripts/ folder:
- detect_triggers.py — Detect escalation triggers in text
- assess_intensity.py — Assess conflict intensity level
- generate_stabilizing.py — Generate calming language
- tone_analyze.py — Analyze tone for aggression indicators
# Detect triggers
# Assess intensity
# Generate calming responseThis skill provides de-escalation techniques for conflict resolution. For immediate danger, always contact emergency services (911).
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

