# Walk Away

> Prepare one negotiation before the room - a B2B deal, a raise, a recruitment close, collective bargaining or a hard
> 1:1 - then rehearse the reply to what the other side says. Write the mandate for every axis on the table (entry,
> ideal, objective, escalation point, walk-away); the browser checks it for gaps and order, suggests a precise anchor,
> works out a concession ladder that shrinks step by step (and the Ackerman steps for a buyer), the zone of agreement
> when their limit is known, reads what they have said, and flags any planned line that asks why, splits the
> difference or says your walk-away out loud. Metered runs write the plan - their stakes, the anchor line, a ladder of
> trades, calibrated questions, an accusation audit, the shared objective, the pause, a pre-mortem and the back-brief -
> and coach the reply to what they just said. Built on the SkillSafe skill @samber/influence-and-negotiation, from
> samber/cc-skills (MIT), by Samuel Berthe. It never invents a competitor's price, a deadline or a market figure.

https://walk-away.skillsafe.ai/

## What is free
- The six intake answers: stage, stakes, counterparty, their exact words, your authority, your walk-away.
- The Mandascan per axis - entry, ideal, objective, escalation, rupture - with missing points, wrong order and a flat entry-to-objective flagged.
- A precise anchor in place of a round entry; the zone of agreement when their limit is known, or the warning that there is none.
- A concession ladder that concedes 50, 30 then 20 percent of the gap from entry to objective; the Ackerman steps (65, 85, 95, 100 percent of target) for a buyer's price axis.
- A read of their words: the four root objections (price, timing, authority, no need) and the signals (a competing alternative, time pressure, emotion, identity).
- A linter for your planned lines: why-questions, yes-or-no questions, splitting the difference, and any private figure (an escalation point or a walk-away).
- Exports: the private mandate (.csv), everything worked out (.json); after a run, a one-page card for the room with no private figures.

## The metered lanes
Two tasks on gpt-terra, metered per run, routed by `task`. Input string fields: `task` (prepare or respond), `situation`, `mandate`, `said` (respond only) and `facts`.
- `prepare` replies with `summary`, `gaps`, `pos` (position, objective and stake per axis), `anchor`, `ladder` (give and get), `questions`, `labels`, `sco`, `pause`, `premortem`, `back_brief` and `caveats`.
- `respond` replies with `summary`, `triage` (emotional, belief, bad-faith, identity, tactical), `root`, `read`, `reply` (label, question, line), `silence`, `trade`, `avoid`, `mandate_check` and `caveats`.
The page holds every reply to the skill's discipline: questions that start with What or How, labels that name the emotion, every concession traded, nothing split, no private figure in a line meant for the other side, and no figure the input does not give.

## Links
- API tutorial: https://walk-away.skillsafe.ai/api.html
- Source skill: https://skillsafe.ai/skill/@samber/influence-and-negotiation