7-step-negotiation

name	7-step-negotiation
description	Walks a category manager or sourcing manager through the classic 7-step negotiation process for a specific deal — Prepare, Set Targets, Build the Concession Plan, Open, Bargain, Close, Measure. Each step produces an artifact (target sheet, opening position, concession log, post-mortem). Persists state in the matter / category folder so a paused negotiation can resume mid-step. Use when the user says "negotiation plan", "walk me through the negotiation", "7-step", "structured negotiation", "I'm negotiating with X, prep me", or wants a step-by-step process rather than a single research packet (which is /negotiation-report).
argument-hint	<counterparty> [--step 1-7] [--category <category>]

/vallor:7-step-negotiation

Purpose

A negotiation goes well when the prep is done before the first meeting. This skill structures the prep, the bargaining, and the post-mortem into seven discrete steps. Each step produces an artifact. State persists — pick up where you left off.

Load context

1. Read ~/.claude/plugins/config/vallor/CLAUDE.md → ## Negotiation posture. Apply the user's default approach, concession discipline, walk-away authority.
2. If --category is set, load that category's strategy from category-strategies/ if it exists.
3. Load the Contract Brief if one exists for this counterparty.
4. Apply the /vallor:design skill — all artifacts use the plugin's monotone palette.

Ask the user for context (optional but encouraged)

Use the AskUserQuestion tool to gather optional context that materially changes the negotiation plan. The goal is a tailored, higher-quality output, not interrogation.

Up front (high-level, before heavy research): Ask 1-3 short questions about anything you cannot infer from the inputs, the practice profile, or the category strategy. Examples for this skill:

• Posture for THIS deal: defensive (protect status quo) / moderate / assertive / aggressive — and why
• Target outcome: savings band, term length, risk reallocation, exit rights, all of the above
• Counterparty history: first deal, renewal, contested renewal, prior dispute
• Walk-away authority confirmed at what threshold
• Hard must-haves and known give-ups going in
• Internal stakeholders who must sign off (and who is loudest)

Mid-flight (after doing some work): After the Step 1 knowledge pack, after market research, or once you see the counterparty's pressure points, you will surface concrete decisions the user has not settled. Stop and ask before guessing on:

• Ambiguities in the brief or category strategy
• Trade-offs between the priorities (e.g., price vs term, SLA vs liability)
• Lever ordering — what do we open with, what do we save
• Information asymmetry — do we reveal X to gain Y

Rules:

• Every question must offer a reasonable default the user can pick to skip ahead.
• Ask follow-ups when answers reveal more depth or a non-default choice.
• Cap a single round at ~3 questions. Batch related questions.
• Skip entirely when the user has signaled a fast / default run or said "don't ask, just do it."

State file

Per negotiation, write to: ~/.claude/plugins/config/vallor/negotiations/<YYYY-MM-DD>-<counterparty>-<category>/state.yaml

The state file tracks:

• Counterparty, category, deal type
• Current step (1-7)
• Artifacts produced (paths)
• Last update timestamp
• Sessions held (with date and summary)

On invocation, the skill reads the state file (if it exists) and resumes at the current step. Pass --step N to jump to a specific step.

The seven steps

Step 1 — Prepare

Goal: know more than the other side.

Build a knowledge pack:

• Counterparty financial health (last 4 quarters earnings, recent M&A, leadership changes, analyst ratings)
• Their pressure points (what's their stock doing, are they up for renewal with another big customer, are they expanding into our region)
• Our spend history with them (Pareto position, payment terms, performance history)
• Competitive alternatives (4-6 named, with rough price differentials)
• Market state (commodity index if applicable, regulatory shifts, market growth)
• Our internal stakeholders (who cares, who has authority, who has to live with the deal afterward)

Artifact: prepare.md — a structured knowledge pack.

Use web search for external research. Tag every claim with source.

Output of Step 1. Knowledge pack saved. Decision tree:

1. Move to Step 2 — set targets.
2. Build the full negotiation packet — /vallor:negotiation-report produces a richer artifact with charts and email drafts.
3. Stop here — knowledge pack is reusable across negotiations.

Step 2 — Set Targets

Goal: know what success looks like before you walk into the room.

For each issue:

Issue	Aspirational (open here)	Target (would close at)	Walk-away (won't sign below)

Cover at minimum:

• Unit price / total spend
• Term length
• Payment terms
• Volume commitments / minimums
• SLA / performance metrics
• Termination rights
• Liability cap (handoff to legal playbook positions)
• Renewal mechanics
• Audit rights

Each target is justified — "Aspirational price is 12% below current because the benchmark shows competitors at 8-15% lower for similar volume." Aspirational without justification is a wish, not a target.

Walk-away authority check. If any walk-away exceeds the user's authority per the practice profile's escalation matrix, surface it:

"Your walk-away on price ($X reduction) is within your authority ($Y threshold), but the walk-away on liability is not — that needs GC sign-off before we open. Want me to draft the escalation note?"

Artifact: targets.md — the target sheet.

Step 3 — Build the Concession Plan

Goal: know what you'll trade before you sit down.

For each issue, define:

• The opening position
• The reciprocal asks the user will demand for any concession
• The order in which concessions happen (less-valuable first)
• The "package deal" — bundles of issues that move together

A good concession plan answers: "If they push hardest on price, what's the chip I trade?"

Artifact: concession-plan.md — a matrix of moves and reciprocal asks.

Step 4 — Open

Goal: open at the aspirational position, not the target.

Drafts the opening communication — usually an email or term sheet. Tone is set from the practice profile's Default approach.

Artifact: opening.md — the opening position document or email draft.

After the user sends and gets a response, log the response in session-log.md and move to Step 5.

Step 5 — Bargain (iterative)

Goal: structured back-and-forth, tracked in a concession log.

For each round:

• What did they ask for / push back on?
• What concession plan move applies?
• What did we agree to give? What did we get in return?
• Updated position vs target vs walk-away — are we still in the zone?
• What's the next ask?

Artifact: concession-log.md — chronological log of every move.

A round-end check: "We've now conceded $X total. That's [Y%] of the aspirational delta. We're still [above/at/below] target. Continue or hold?"

When the walk-away line is crossed: stop the user. "We're about to cross the walk-away on [issue]. This is a stop-and-think moment. Three options: (a) hold the line and risk no deal; (b) escalate to [approver from matrix] for a new walk-away; (c) signal we're at our final position and offer one closing sweetener."

Step 6 — Close

Goal: lock the agreement before it drifts.

Produces a deal memo summarizing every agreed term, in plain English, for both sides to sign off on BEFORE the long-form contract is drafted. This is the "this is what we shook on" document that prevents drafting drift.

Artifact: deal-memo.md — every agreed term, both sides sign or initial.

Handoff to drafting: /vallor:contract-prep --brief <Contract Brief> or directly to /vallor:contract-redline-docx if the counterparty is on their paper.

Step 7 — Measure

Goal: post-mortem and feed forward.

Two weeks after signing, capture:

• Did we hit our targets? On which issues did we exceed / miss?
• What did we concede that we didn't need to? Why?
• What did they concede that we didn't expect? Why?
• Counterparty intel that's worth retaining for next time.
• Personal moves and tactics that worked / didn't.

Artifact: post-mortem.md — feeds back into category strategy + practice profile (the playbook-monitor agent picks up patterns over time).

The post-mortem is the single highest-leverage step. Skipping it means the same mistake gets made next quarter.

Cross-step rules

• Every artifact begins with the work-product header from the practice profile (PRIVILEGED & CONFIDENTIAL for legal contexts; internal-only for pure procurement).
• Every artifact ends with a single-line summary that's safe to forward internally.
• Don't number the steps in the artifacts (e.g., don't title it "Step 5 — Bargain"). Title them by the artifact ("Bargain Log — Acme MSA Renewal"). Save the step numbering for the state file.

Decision tree at any step

1. Move to the next step.
2. Re-do the current step with updated inputs.
3. Jump to the post-mortem — applies when a negotiation is paused or abandoned; capture the learning anyway.
4. Generate the full negotiation packet — /vallor:negotiation-report produces the parallel research-heavy artifact for the same deal.
5. Stop here — state is persisted; resume by re-running the skill.

Examples

/vallor:7-step-negotiation Acme

/vallor:7-step-negotiation Megacorp --step 5

/vallor:7-step-negotiation Acme --category "IT Hardware Maintenance"