> ## Documentation Index
> Fetch the complete documentation index at: https://docs.hud.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Harbor coding tasks

> Bring Harbor tasks and the runs they recorded into HUD, find the broken ones with QA agents, and run the fixed tasks on HUD at scale.

For teams with coding tasks in [Harbor](https://harborframework.com) format: task folders with an
instruction, an environment, and a verifier. HUD reads them as they are. Import the tasks and the
runs Harbor already recorded, let QA agents find the tasks whose verifiers can't be trusted, fix
them, and run the fixed set on any model on HUD's hosted infrastructure.

## Two ways in

| | Import | Adapt and deploy |
| - | - | - |
| **What you get** | Your tasks and recorded runs in a taskset, to browse, compare, and review with QA. | HUD environments and tasks that run on HUD, on any model. |
| **What runs** | Nothing. Imports are free. | Every evaluation you start, billed as usual. |
| **How** | [Harbor import](/platform/harbor-import) from the web app, `hud harbor import`, or the REST API. | The SDK's [Harbor adapter](/v6/experimental/harbor#adapt-harbor-tasks), then `hud deploy` and `hud sync tasks`. |

Most teams start with an import to see where the benchmark stands, then adapt the tasks to run
them on HUD.

## Find the broken tasks

<Steps>
  <Step title="Import tasks and runs">
    Zip your task folders and job folders, and import them into a taskset. Each task folder becomes
    a task, each job an evaluation job, and each finished trial a trace with its trajectory and
    verifier reward. See [Harbor import](/platform/harbor-import).
  </Step>

  <Step title="Run QA on the runs">
    In the taskset's task grid, select the tasks and select **Run QA** to review their traces. See
    [run QA](/platform/run-qa#from-a-taskset).
  </Step>

  <Step title="Read what QA found">
    Each check points at the steps and files involved. See [read results](/platform/qa-results).
  </Step>

  <Step title="Fix and import again">
    Change the task folder and import it again. A task whose content changed moves to a new version,
    and runs you import later link to the version they ran. See
    [import again](/platform/harbor-import#import-again).
  </Step>
</Steps>

What each trace check finds in coding tasks:

| QA agent | Finds |
| - | - |
| **Reward Hacking** | Agents that pass by editing tests, weakening the verifier, or special-casing the checked inputs. |
| **False Positive** | Verifiers that pass broken work: tests that only check a file exists or never exercise the change. |
| **False Negative** | Correct solutions the verifier rejects, for example over formatting or a stale expected value. |
| **Prompt Alignment** | Requirements the tests enforce but `instruction.md` never states. |
| **Failure Analysis** | Whether each failure comes from the agent, the environment, the verifier, or the task. |

## Run the tasks on HUD

The SDK adapts Harbor task folders into HUD environments and tasks, keeping the Harbor environment
and verifier:

```bash theme={"dark"}
hud deploy .hud-adapt/<environment>/env.py --runtime-config .hud-adapt/<environment>/runtime.json
hud sync tasks my-harbor-tasks .hud-adapt/<environment>/tasks.json
```

See [adapt Harbor tasks](/v6/experimental/harbor#adapt-harbor-tasks) for the full flow. Once
deployed, run the taskset on any [model](/platform/models) from the web app or with `hud eval`, as
many times as you need. Turn on [automatic QA](/platform/run-qa#automatically) so every new trace is
reviewed when it completes, and use [file tracking](/platform/file-tracking) to see the code the
agent changed at each step.

<Columns cols={3}>
  <Card title="Harbor tasks" icon="ship" href="/platform/harbor-import">
    Import, adapt, and export, with the zip layout and every limit.
  </Card>

  <Card title="QA agents" icon="https://mintcdn.com/hud-f5fd7c15/9ozVUWR-rlBrAGzN/platform/icons/qa-agent.svg?fit=max&auto=format&n=9ozVUWR-rlBrAGzN&q=85&s=c2c263fc6b61484a24b706ba5c61df0b" href="/platform/qa-agents" width="32" height="32" data-path="platform/icons/qa-agent.svg">
    What each check reads and decides.
  </Card>

  <Card title="Harbor interoperability" icon="code" href="/v6/experimental/harbor">
    Adapt, import, and export from the SDK.
  </Card>
</Columns>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.