Metric Setter

Who the node is. The id is the key the DOT pins.

id

metric-setter

name

Metric Setter

type

human-input

phases

Planning

What it does, and the prose the agent is handed when the graph runs.

action21 words

Take the measure from the person who will judge the run, and put it where only the monitor can read it.

spec124 words

Ask the person accountable for this run how they will know the objective was met, and record their answer as conditions. Each condition should be one a reader holding the run's state and that condition alone could settle. Press for conditions that would be false if the objective were missed: a measure that is satisfied by almost any run is not a measure. Record what would count as failure as well as what would count as success, because a run that ends ambiguously is the case this node exists to prevent. Emit them on success_metric. This is the only node in the graph that produces them and exactly one node reads them; do not restate any of it anywhere a working node could see.in full above

model

whatever the graph supplies

agent

not named

skill

not named

tools

Human review

mcp

none

params

none

What arrives, what leaves, which nodes it expects to hear from, and what may not.

inputs

none

outputs

success_metric : acceptance-criteria

dependencies

none

cannot

no type is refused

will_not

state the measure anywhere a working node could read it, accept a measure that almost any run would satisfy

The keys the static analysis reads. Nothing here instructs the agent.

risk_markers

none

notes25 words

A second human-input root, deliberately separate from goal-setter. Splitting them is what lets the objective reach the planner while the measure reaches only the monitor.

The card's own version, and who wrote it.

version

1.0.0

author

autogen

provenance

not stated