Does this preview call an AI model?

The explicit metadata-only run preview does not start a model session.

Is --executor dry-run the same thing?

No. An executor is part of the normal run loop. Use the explicit --dry-run flag for the metadata-only preview.

Will the preview prove my tests pass?

No. It does not execute verification profiles or establish readiness.

How to preview work with the Liquid Brain harness CLI

Short answer

Use plan to inspect work selection and run --dry-run for the explicit metadata-only run preview. A preview explains selection without starting a model session; it does not establish execution readiness.

Before you start

Run these commands from the installed harness directory with a configuration that points to your intended workspace. Replace demo-config.json with your configuration path. The current hosted-write path requires Python 3.12 or newer and the pinned installed Brain tooling; the Windows Desk source candidate has its own CPython 3.14 AMD64 requirement.

Read the selection before running work

The plan lists startable cards and why others were not picked. Priority, current state, dependencies and scope affect eligibility. A blocked or decision-waiting card should not silently become an automatic work item. Review the workspace path printed by the command before drawing conclusions.

Commands

From the harness directory:

python pm.py --config "demo-config.json" plan
python pm.py run --config "demo-config.json" --dry-run

The first command reports selection. The second uses the explicit read-only metadata preview branch. Neither command is a request to execute the selected card.

Do not confuse the flag with the executor

The --dry-run flag is distinct from an executor named dry-run. The executor can be used inside the normal harness loop; it is not a promise that the surrounding loop leaves card state unchanged. Do not substitute plain run or --executor dry-run for the explicit preview command in this guide.

Steps

Open a terminal in the installed harness directory.
Identify the configuration for the workspace you intend to inspect.
Run python pm.py --config "demo-config.json" plan using your configuration path.
Read the selected cards and reasons for skipped work.
Run python pm.py run --config "demo-config.json" --dry-run to request the explicit metadata preview.
Review the output; do not interpret preview success as permission or readiness to execute.

Expected result and troubleshooting

Expect a selection or explanation that nothing is startable. If no workspace is configured, supply the correct configuration rather than guessing a path. If a dependency or scope prevents selection, review that card record. A successful preview does not test active protection, execution grants, provider setup or verification profiles; those belong to execution readiness and the real run.