---

name: primitives description: "Guides agents using StatevectorSampler, StatevectorEstimator, primitive jobs, PUBs, and primitive result containers." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

Qiskit primitives workflows

Use this sub-skill when the task is about Qiskit's primitive interfaces: sampling circuits, estimating expectation values, binding parameter sweeps, and reading primitive result containers.

Read next

• references/workflows.md for sampler/estimator recipes and result-access patterns.
• references/troubleshooting.md for data-shape mistakes, missing measurements, and result-indexing errors.
• ../../references/module-map.md for the package map.
• ../../scripts/check_qiskit_environment.py --sections primitives for a source-free primitive smoke check.

Include here

• StatevectorSampler, StatevectorEstimator, PUBs, and run() usage.
• Primitive job/result container classes such as PrimitiveResult, PubResult, SamplerPub, EstimatorPub, DataBin, and BitArray.
• Parameter sweeps, shots, and precision settings for the reference implementations.
• Interpreting measured counts, expectation values, and metadata from primitive results.

Exclude or route elsewhere

• Backend classes, provider implementation, and fake backends belong in ../providers/SKILL.md.
• Abstract circuit construction belongs in ../circuit/SKILL.md.
• Serialization of circuits belongs in ../serialization/SKILL.md.
• State/operator mathematics belongs in ../quantum-info/SKILL.md.
• Output rendering belongs in ../visualization/SKILL.md.

Default route

Start here when the user wants to sample from a circuit, estimate an observable, or understand the nested data containers returned by a primitive job. If the problem starts with a backend class rather than a sampler or estimator, route to providers first.

What to remember

• Sampler results are organized around measured bit arrays; estimator results are organized around expectation values.
• PUBs may be provided directly or via the convenience tuples accepted by run().
• The shape of the result container mirrors the shape of the observable and parameter sweeps.
• Measured circuits are usually required for sampling, while estimators work with unmeasured circuits and observables.