Imported from 0xUrsanomics/utopia-os (skills-shared/negotiation/SKILL.md). Install upstream with npx skills add 0xUrsanomics/utopia-os --skill negotiation. Copyright stays with the author.

Negotiation Skill

Tactical negotiation support. Game theory, emotional regulation, structured analysis. Not an academic framework. Built from live deal experience.

Source & Pairs with

• Your deal-pipeline state file. the live state of every deal (read BEFORE any negotiation analysis to avoid stale-recall on price/scope)
• Your company-context file. your pricing tiers, service-line boundaries, and org context
• Your pushback-discipline feedback. never relay a kickback mid-deal; keep pricing floors
• Sister skills: a pipeline-ops skill (qualification), an outreach-drafter (drafts the email side), skills-shared/sycophancy-guard/SKILL.md (catches drift into compliance when pushback is correct)

Example output (counter-offer analysis)

## Situation
Counterparty: Example Corp HQ (their partnerships team)
Their offer: $15K one-time sponsor, no retainer
Our ask: $33K standard + $5K/mo retainer (3 month minimum)
BATNA: walk; another activation on the calendar is more leverage-rich

## Game-theory read
- They've shown an HQ-budget ceiling pattern (evidence: 2x retainer-refusal in 6 weeks)
- Their move isn't disrespect. it's a structural constraint from HQ
- Our walk-away cost: ~$15K one-time event revenue
- Their walk-away cost: lose a market-specific intel feed

## EV table
| option | EV | downside | upside |
|---|---|---|---|
| Accept $15K | +$15K, no retainer | sets precedent | quick close |
| Counter $25K + 1-mo retainer | +$30K + retainer signal | 60% accept | proves the pattern broken |
| Walk | $0 | lose event | preserves the pricing floor |

## Recommendation
Counter $25K + 1-mo retainer. EV-positive vs walk, preserves the pricing floor, surfaces a real constraint signal.

When This Activates

• The user is in a live negotiation (forwarding counterparty messages, asking "what should I say")
• Pricing pushback from a client or partner
• Contract term disputes
• Any back-and-forth where money, scope, or commitments are being discussed
• The user asks for BATNA, EV analysis, or a walk-away assessment

Phase 0: Emotional Check (ALWAYS FIRST)

Before drafting ANY response for the user to send:

Scan the user's messages for heat signals:

• Profanity directed at the counterparty
• "stupid", "idiot", "wtf", "pissed", "annoyed", "can't believe"
• ALL CAPS
• Rapid-fire short messages (frustration cadence)
• "just tell him [aggressive thing]"

If heat detected:

1. Acknowledge it: "you're pissed, I get it"
2. Do NOT draft the reply yet
3. Ask: "want to vent first or go straight to the tactical response?"
4. If they say tactical, draft a COOLED version. never match their heat in the outbound message.
5. Flag inline: "your draft would have been [X], I wrote [Y] instead because [reason]"

Why: angry messages in negotiations are almost always regretted. the skill's job is to be the 5-second delay between emotion and the send button. this is the single highest-value function.

Phase 1: Situation Map

Before any analysis, extract these from context:

Present as a compact table. Ask the user to confirm or correct.

Phase 2: Pattern Detection

Scan the conversation for known negotiation tactics:

Tactics the counterparty might be using

Tactics you should use

Phase 3: Concession Tracking

Log every price/term move by both sides in a table:

Rules:

• Never make more than 2 concessions without getting one back
• Each concession should be smaller than the last (a diminishing pattern signals the floor)
• If you've made 3+ moves and they've made 0, STOP. say: "I've moved three times. What can you move on?"
• Track cumulative concession as a % of the original ask. Over 40% cumulative = you've given too much

Phase 4: EV + Bayesian Analysis

Run this whenever the user asks or at major decision points.

Bayesian Table

Start with a prior P(deal closes on acceptable terms) based on:

• Lead source quality (warm intro = higher prior)
• Counterparty engagement level
• Industry norms for this deal type

Update with each new signal:

| Signal | Direction | Weight | P before | P after |
|--------|-----------|--------|----------|---------|
| [evidence] | +/- | weak/moderate/strong | X% | Y% |

EV Table

For each possible outcome:

| Scenario | P(%) | Revenue | Time cost | Opp cost | Net EV |
|----------|------|---------|-----------|----------|--------|
| S1: deal on your terms | X | $ | hrs | $ | $ |
| S2: deal on compromise | X | $ | hrs | $ | $ |
| S3: they walk | X | $0 | sunk | freed | $ |
| S4: you walk | X | $0 | sunk | freed | $ |
| S5: you accept bad terms | X | $ | hrs | $ | $ |

Weighted EV = sum of (P * Net EV)

Decision rule: if the weighted EV of walking > the weighted EV of continuing to negotiate, recommend walk.

Phase 5: Counter-Offer Drafting

When drafting a response for the user:

1. Match the channel (chat = short and direct, email = structured, call = talking points)
2. Match the user's register (if the user writes casual, draft casual. if the counterparty is formal, draft formal for their consumption.)
3. One ask per message. don't bundle multiple concessions or demands.
4. Always end with a question or a clear next step. never end with a statement that lets the conversation die.
5. Include a deadline when appropriate. "let me know by [date]" creates urgency.
6. Offer exactly two options when possible. binary choices are easier to decide on than open-ended asks.
7. De-slop every draft. run it through the anti-ai-slop filter. zero em dashes, zero filler.

Response templates by situation

They say "too expensive":

"[acknowledge]. I can adjust the structure. here's what the [lower tier] looks like: [specifics]. the scope stays the same / adjusts to [X]. which works better for your budget?"

They want free work first:

"I get the logic. but [deliverable] requires [real costs]. who covers those during the trial? I need a minimum operational commitment to execute. [specific floor number]."

They use "believe in yourself" / emotional framing:

"this isn't about belief. [evidence of work already done]. this is about operational cost. [specific costs]. how do you see that working for both sides?"

They threaten to walk:

"I understand. I don't want to push terms that don't work for you. if things change, my door's open." (do NOT chase. do NOT counter. silence is the response.)

They go silent after your offer:

Wait 3-5 business days. then ONE follow-up: "hey [name], checking in on [topic]. offer stands until [deadline]. let me know either way." (ONE follow-up max. after that, radio silence until they come to you.)

Phase 6: Walk-Away Framework

Auto-flag walk-away when ANY of these are true:

1. Weighted EV of the deal < $0
2. You've made 3+ concessions and they've made 0
3. They're asking for free work with no guaranteed payment
4. The deal would set a precedent that damages future pricing
5. Your emotional state has been negative for 3+ exchanges about this deal
6. Opportunity cost: time spent here blocks a higher-EV deal
7. They've disrespected your time or expertise repeatedly

Walk-away is not failure. it's a strategic decision that preserves:

• Your pricing integrity for future deals
• Your time for higher-EV opportunities
• Your emotional health
• The possibility of re-engagement on better terms later (they often come back)

Walk-away message format: Keep it short. grateful, firm, door open. never explain why you're walking (that gives them negotiation ammo). just state the outcome.

Phase 7: Post-Negotiation Debrief

After any negotiation concludes (deal closed OR walked):

1. Log the concession table to Decisions.md
2. Grade each move (A-F)
3. Identify the weakest link
4. Extract a reusable lesson for Learnings.md
5. If it was a new tactic encountered, add it to the tactics table above
6. Update the deal page in your knowledge graph with the outcome

Game Theory Quick Reference

Interaction with Other Skills

What This Skill Does NOT Do

• Does not negotiate on behalf of the user without approval (all outbound messages are CONFIRM)
• Does not guarantee outcomes. negotiation is probabilistic.
• Does not replace the user's judgment. it surfaces analysis, the user decides.
• Does not make the user confrontational for the sake of it. cooperation is usually higher EV than aggression.