Conduct

The checks cannot tell you what an agent will do with a task. Conduct runs it.

Run: `run_harbor_tasks`

run_harbor_tasks {"repo": "acme/tasks", "ref": "main",
                  "paths": ["tasks/fix-rounding", "tasks/fix-tax"],
                  "mode": "conduct", "models": ["google/gemini-3.8-flash"],
                  "harness": "terminus-2", "k": 8, "reasoning_effort": "medium"}

Argument	Meaning
`mode`	`conduct` (default), `validate`, or `full`: the validate stages, then one rollout with the first model
`models`	Catalog model ids, required for `conduct` and `full`. The ones your tier may run are this field’s enum
`harness`	The agent the model runs under. Default `terminus-2`
`k`	Trials per task and model
`reasoning_effort`	Required. Passed verbatim to the harness’s own setting; Kennet has no default
`max_output_tokens`	The output limit per model call. 0 keeps the harness default, recorded as such
`analyze`	Review each trial’s trajectory as it finishes. Default true
`max_attempts`	Re-runs of a trial that failed on infrastructure. Default 3
`agent_timeout_s`	Cap the agent below the task’s own `[agent] timeout_sec`, for a fast screen; never above it. 0 keeps the task’s limit
`verifier_judge_model`	The verifier’s judge, passed as `JUDGE_MODEL`. Empty keeps the task’s own
`harbor`	Harbor’s own job config (`harbor run --config`), any part of it, in Harbor’s field names. A field Kennet sets (the task, k, timeouts, the model, the network) is refused with why
`user_model`	The model of a simulated user that `harbor.user_agent` names

Every run records the repository and the commit it ran at.

Models and cost

A model that cannot be served refuses the launch before any trial starts, and nothing is spent. A conduct run names its models; one with none is refused. harbor_rollouts reports each trial’s cost and the cost per solve per model and harness.

What lands where

<task>/harbor/                         the bundle as it last ran
<task>/eval/<run id>/<model slug>/     the run: each trial's result.json, logs, trajectory
                                       rollouts.jsonl, one row per trial

Failures are typed

Each trial ends as one of these, read from what Harbor recorded at the source. The outcome says whether the grader ran: a trial with a verifier result is graded and carries its reward, and it counts as solved where that reward is 1.0 or more (pass@k and solve rates read it so). Only a trial with no verifier result gets an error outcome.

Outcome	Meaning	In the score
`graded`	The verifier graded the trial, whatever its reward	its reward
`agent_context`	The agent overran its context window, and no verifier ran	0
`agent_timeout`	The agent spent its time budget, and no verifier ran	0
`infra_error`	Sandbox, image, network or provider failure	not counted
`upstream_error`	The model call failed after retries	not counted
`ungradable`	The verifier could not grade: a bundle or image defect	not counted
`missing_result`	No reward and no recognised cause	not counted

Infrastructure is never scored as the model’s 0. It is counted beside the score, and retryable outcomes run again up to max_attempts. When a failure would hit every call alike (the model service down, credits out), the launch stops the calls not yet rolling out and marks them stopped. The task’s result names the failure and carries rerun: the one call that runs exactly the stopped work again.

Follow and stop

Every run is a task. get_task shows each call’s phase and any failure as it happens. cancel_task cancels every call and stops its sandboxes.

Getting started

Taskforge

Twin API

Deliveries API

Reference

Run: `run_harbor_tasks`

Models and cost

What lands where

Failures are typed

Follow and stop

​Run: run_harbor_tasks

​Models and cost

​What lands where

​Failures are typed

​Follow and stop

Run: `run_harbor_tasks`

Models and cost

What lands where

Failures are typed

Follow and stop