> ## Documentation Index
> Fetch the complete documentation index at: https://zuno-fb55ec99.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# zuno plan

> Run the four-agent debate and produce a rebalance plan.

```
recommend what I should do with this position
```

The headline workflow. When the AXL mesh is online, the CLI sends the
request through Scout → Strategist ↔ Critic → (Arbiter on deadlock)
and returns a structured, persisted plan with the full debate
transcript attached. If the mesh is not online, an in-process
orchestrator runs the same four handlers; if no LLM is available, the
deterministic `recommendPlan` math runs as a final fallback.

## What happens

1. CLI sends a `flow_start` envelope to **Scout**.
2. **Scout** reads the position + pool tick state, loads vol / gas /
   yield, classifies the regime (`ranging` / `trending` / `volatile` /
   `stressed`), and forwards `context_observed` to **Strategist**.
3. **Strategist** proposes 2-5 candidate ranges as
   `(widthMultiplier, centerOffsetTicks)` tuples (snapped to
   `tickSpacing`), each with stress + yield numbers. Forwards
   `proposal` to **Critic**.
4. **Critic** runs deterministic stress (1×, 2×, 3× vol), gas/yield
   ratios, and judges each candidate `accept` | `revise` | `veto`
   against the user's risk profile.
   * On `accept`: emits `plan_ready` to the CLI, flow ends.
   * On `revise`: forwards `critique` back to Strategist for round 2.
   * On `veto_all` past `ZUNO_MAX_DEBATE_ROUNDS`: emits `deadlock` to
     **Arbiter**.
5. **Arbiter** (on deadlock) reads the full debate, picks one candidate
   from the latest proposal, and emits `plan_ready` to the CLI.
6. CLI prints the result and writes `~/.zuno/plans/<planId>.json` with
   the transcript baked into `risk.reasons`.

The CLI streams `agent_thought` envelopes throughout, rendering each
turn live under the recommendation card:

```
  ◇ scout       regime: ranging · vol 110bps · gas 14gwei · in range
  ◇ strategist  3 candidates · lean tighter, vol is sleeping
  ◇ critic      reject A: 18h buffer is fine now, vol mean-reverts
  ◇ strategist  A′: ±2.6% · 28h buffer · $3.60/d expected
  ◇ critic      accept A′ - converged
```

## Example output

```
$ recommend what I should do with this position

  position    WETH / USDC 0.05%
  range       2,190.79 to 2,421.19   out of range
  current     2,438.19

  ────────────────────────────────────────────────────────

  recommended 2,273.36 to 2,614.97   shift
  rejected    2,361.41 to 2,519.99   tighten at 2× vol
  reason      28h base buffer · survives 2× vol · gas/yield 0.71x
  decided by  critic · debate converged

  confidence  0.82  approve_with_caution

  plan id     plan_dd4f9e0b6806
```

## Risk profile

Set `ZUNO_RISK_PROFILE` to bias the Critic's floors and the Arbiter's
tiebreak:

| profile        | min base buffer | max gas/yield | tiebreak                                  |
| -------------- | --------------- | ------------- | ----------------------------------------- |
| `conservative` | 36h             | 1.2x          | largest 2× vol buffer wins                |
| `balanced`     | 24h             | 1.6x          | best buffer × yield balance               |
| `aggressive`   | 14h             | 2.4x          | highest yield among non-vetoed candidates |

## Confidence labels

| label                  | meaning                                                  |
| ---------------------- | -------------------------------------------------------- |
| `approve`              | converged with strong buffer                             |
| `approve_with_caution` | converged but buffer is tight, or arbiter broke deadlock |
| `reject`               | every candidate failed stress at the active risk profile |

## Replaying the debate

```
explain recommendation
```

Renders `risk.reasons`, which includes every `[scout]` / `[strategist]`
/ `[critic]` / `[arbiter]` line from the original flow. The transcript
is persisted with the plan so you can read the reasoning later.