---

name: circuits-devices description: "Build and debug PennyLane QNodes, devices, measurements, shots, drawing, and device-test CLI workflows." metadata: disco-role: operating disable-model-invocation: true license: Apache 2.0

Circuits, QNodes, devices, and measurements

Use this sub-skill when the task asks how to create, execute, inspect, or debug PennyLane circuits. It owns QNodes, device selection, measurements, shots, drawing, mid-circuit measurement routing, and the device-test CLI.

Read first

• references/api-reference.md: signatures and core objects for QNode, qnode, device, measurements, drawing, and execute.
• references/workflows.md: recipes for QNode creation, finite shots, drawing, and device-test checks.
• references/troubleshooting.md: device, wire, shot, measurement-shape, and mid-circuit failure modes.
• scripts/qnode_smoke.py: safe CPU QNode smoke script.
• scripts/device_help.sh: helper to display pl-device-test --help when the console script is installed.

Route within this sub-skill

• Basic circuit creation: use qp.device(...), define a quantum function with operations, return measurement processes, and wrap it with @qp.qnode(dev) or qp.QNode(func, dev).
• Device choice: start with default.qubit for ordinary pure-state CPU simulation; use default.mixed for noisy/mixed-state workflows, default.clifford for Clifford circuits, default.tensor for tensor-network style simulation, reference.qubit for reference checks, and null.qubit for dry/resource-like runs.
• Measurements: use qp.expval, qp.var, qp.probs, qp.sample, qp.counts, qp.state, qp.density_matrix, entropy/purity/mutual-info functions, and shadows according to the requested return type.
• Shots: analytic workflows use exact-style returns; sampling/counts require finite shots. Use qp.set_shots or QNode/device update methods when changing shot behavior.
• Drawing/inspection: use qp.draw(qnode)(*args) for text diagrams and qp.draw_mpl(qnode)(*args) for Matplotlib figures.
• Plugin/device validation: use pl-device-test --help first, then choose explicit --device, --shots, --analytic, and --device-kwargs values.

Boundaries

• Operations/templates and circuit transforms are owned by ../operators-transforms/SKILL.md.
• Gradients/interfaces/training are owned by ../gradients-interfaces/SKILL.md.
• Domain modules such as qchem/qcut/resource are owned by ../applications-qchem-resource/SKILL.md.
• Import/export, datasets, logging, and debugging snapshots are owned by ../io-data-logging/SKILL.md, although snapshot measurements can be used inside circuits.
• Source-checkout test/lint policy is owned by ../repo-development/SKILL.md.

Minimal pattern

import pennylane as qp

angles = qp.numpy.array([0.1, 0.2], requires_grad=True)
dev = qp.device("default.qubit", wires=2)

@qp.qnode(dev)
def circuit(params):
    qp.RX(params[0], wires=0)
    qp.CNOT(wires=[0, 1])
    qp.RY(params[1], wires=1)
    return qp.expval(qp.Z(1))

value = circuit(angles)
print(value)
print(qp.draw(circuit)(angles))

Verification cues

A good answer or code change in this area should state the chosen device, wires, shot setting, measurement return shape, interface/diff assumptions if gradients are involved, and a tiny QNode smoke check. For source-checkout changes, native candidates include QNode, device, and measurement tests mirrored under tests/workflow/, tests/devices/, and tests/measurements/.