Interface: StepsUnfinishedRow

Defined in: src/core/agent/assessment/witness.ts:88

The answer came before the active skill's declared steps finished (agentfootprint.skill.steps_unfinished with action: 'accepted' — the one teaching nudge was already spent — or 'cut-short' — a limit forced the answer). Filed by the Route decider while the answer layer is armed, and only for the answer that STANDS: the step judge runs before the evidence gate, so a stop it accepted on a draft the gate then sends back files no row (the event still fires) — the revision is judged again, and at most one row per answer reaches the ledger. The 'nudged' verdict needs no row: the turn went on.

Properties

action

readonly action: "accepted" | "cut-short"

Defined in: src/core/agent/assessment/witness.ts:100

---

iteration

readonly iteration: number

Defined in: src/core/agent/assessment/witness.ts:93

The iteration whose answer the step judge read.

---

kind

readonly kind: "steps-unfinished"

Defined in: src/core/agent/assessment/witness.ts:89

---

remaining

readonly remaining: readonly UnfinishedStep[]

Defined in: src/core/agent/assessment/witness.ts:97

The declared steps not reached, in order — position and tool only.

---

skillId

readonly skillId: string

Defined in: src/core/agent/assessment/witness.ts:95

The skill whose procedure was in progress.

---

total

readonly total: number

Defined in: src/core/agent/assessment/witness.ts:99

How many steps the procedure declares.

---

turn

readonly turn: number

Defined in: src/core/agent/assessment/witness.ts:91

AgentState.turnNumber when the row was filed — the conversation turn.