Deal intelligence skill for offer analysis and counter-proposal drafting. Trigger on `review this offer`, `analyze counter`, `value gap`, `draft counter`, `should I walk`. Apply when reviewing any offer (real estate, compensation, vendor, resource allocation) or designing negotiation analysis behavior in agentic systems. Quantifies value gaps and applies the RADAR protocol.
Scanned 9/1/2026
Install to Claude Code
npx -y skills add rjmurillo/ai-agents --skill negotiation --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Negotiation?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/rjmurillo-negotiation-ai-agents)More formats (shields.io, HTML) on the badges page.
---
name: negotiation
version: 1.0.0
description: Deal intelligence skill for offer analysis and counter-proposal drafting. Trigger on `review this offer`, `analyze counter`, `value gap`, `draft counter`, `should I walk`. Apply when reviewing any offer (real estate, compensation, vendor, resource allocation) or designing negotiation analysis behavior in agentic systems. Quantifies value gaps and applies the RADAR protocol.
license: MIT
metadata:
domains: [negotiation, deal-intelligence, behavioral-influence, agent-design]
type: knowledge
source: Anthropic Project Deal (Dec 2025), Fisher and Ury, Voss, Navarro, Hughes, Cialdini, Galinsky
---
# Negotiation Skill
Codifies deal intelligence behavior. Use when reviewing any offer or
designing how an agentic system should analyze and counter-propose.
The full crystallized skill set lives in `references/skills.md`.
## Triggers
| Phrase | Context |
|--------|---------|
| `review this offer` | Real estate, compensation, vendor, resource allocation |
| `what's the value gap` | Quantification request |
| `draft a counter` | Counter-proposal drafting |
| `should I walk` | Walkaway analysis |
| `anchor first or wait` | Information-asymmetry decision |
Skip when:
- The change is descriptive ("explain how anchoring works") rather than
applied. Use a knowledge response, not the skill.
- The artifact is not an offer or counter (e.g., a status update).
## Process
Apply the RADAR protocol in order. Skipping a step is the most common
failure mode and the rule against it is load-bearing.
### Phase 1: Read
Extract every term in the offer. List anchors, hedging, urgency
signals, omissions. Do not interpret yet.
### Phase 2: Analyze
Map ZOPA, BATNA, information asymmetry, value gap. Quantify the gap
in dollars before any qualitative language. See
`references/skills.md` Skill-Negotiation-001.
### Phase 3: Design
Draft PCP framing (Perception, Context, Permission) before producing
a number. Whoever defines the frame controls the negotiation. See
Skill-Negotiation-003.
### Phase 4: Assess
Run the invisible-disadvantage check. Compare risk of countering vs.
accepting. Apply the anchor-first decision rule
(Skill-Negotiation-008) and BATNA discipline
(Skill-Negotiation-009).
### Phase 5: Review
Produce a DRAFT counter with explicit human approval gate. Specific
closing pattern from Skill-Negotiation-010 applies once verbal
agreement is reached.
## Verification
Output is acceptable when ALL of the following hold:
- [ ] Offer net value, achievable value, and gap are quantified in
dollars (Skill-Negotiation-001 implementation block).
- [ ] At least one written-signal observation from
Skill-Negotiation-007 is cited if the offer arrived in writing.
- [ ] Counter-proposal carries PCP framing in the output, not just a
number (Skill-Negotiation-003).
- [ ] BATNA worksheet (Skill-Negotiation-009) is filled before the
first counter goes out.
- [ ] Bundling rule (Skill-Negotiation-005) is honored: no
single-dimension concession appears in the counter.
- [ ] Final output is marked DRAFT and waits for explicit human
approval before sending.
- [ ] If routed through an agentic system, the negotiation analysis
runs on a senior-tier model (Skill-Negotiation-006).
## Anti-Patterns
- **Reactive countering**: producing a number before completing
Phase 1 (Read) and Phase 2 (Analyze). Frame is then defined by the
other party.
- **Single-dimension concessions**: trading price without extracting
value on another dimension. Drains leverage item by item.
- **Matching urgency**: accepting tactical deadlines as if they were
genuine. Slow down and require external justification.
- **Mid-negotiation BATNA drift**: relaxing the walkaway under
sunk-cost pressure. Pre-commit in writing; require 24h delay for
any adjustment.
- **Routing to a junior model**: model capability gap is the only
variable that consistently changes negotiation outcomes (Project
Deal). Prompt style does not compensate. See
Skill-Negotiation-006.
- **Verbal-only close**: leaves the deal exposed to remorse and
selective memory. Always confirm in writing within 1 hour
(Skill-Negotiation-010).
## Extension Points
- **New domain (insurance, M&A, etc.)**: add a domain-specific
reference under `references/` mapping the domain's leverage points
onto the 10 skills. Do not edit the canonical skill list.
- **Behavioral framework swaps**: the PCP framing,
comfort/discomfort signal table, and BATNA worksheet are pluggable.
Replace via a sibling reference if a different model fits the
domain better.
- **Agent integration**: when wiring the skill into an agentic
system, include `model_tier: opus` (or equivalent senior tier) in
the agent template.
## References
- `references/skills.md`: full canonical 10 crystallized skills
with implementation patterns, evidence citations, and tags.
- Source citations: Fisher and Ury _Getting to Yes_, Voss _Never Split
the Difference_, Navarro _What Every BODY Is Saying_, Hughes
_Six-Minute X-Ray_, Cialdini _Influence_, Galinsky/Mussweiler/Chen
anchoring research, Anthropic Project Deal (Dec 2025; internal
research, data not publicly available for audit; specific dollar
values illustrative).
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!