Guide · Agent skill
How to use the QuTiP Quantum Simulation Auditor agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The QuTiP skill is a packaged set of instructions, reference notes and small Python helper scripts that an AI agent can load when you ask for quantum-system simulations. It targets QuTiP 5 and is aimed at physics students, researchers and developers who need simulation results whose assumptions, dimensions and numerical settings are written down. After this guide you will know what the skill changes about the agent's behaviour, how to install it, how to phrase requests so it activates, and what to check before trusting the output.
QuTiP Quantum Simulation AuditorK-Dense-AI/scientific-agent-skills · MIT Download the skill Quick start
- 1Download and unzip. Download the skill ZIP from this page and unzip it to get the skill folder.
- 2Place the folder. Copy it to ~/.claude/skills/ for personal use or .claude/skills/ inside a project, or upload it under Settings → Capabilities → Skills in the Claude apps.
- 3Create the environment. Run uv venv --python 3.11 and install qutip==5.3.1 with uv pip install.
- 4Ask for a simulation. Describe the physical system, the solver need and the outputs you want, stating that you need explicit units and convergence checks.
- 5Review the audit. Check the recorded units, approximations, state validity and convergence sweep before using any result.
What the skill makes the agent do
Without the skill, an agent may write a quick QuTiP script. With the skill, it follows a fixed model contract: it records units, subsystem ordering, approximations and numerical settings before solving, chooses a solver that matches the physics, checks state validity and runs convergence sweeps.
The skill covers deterministic, trajectory, steady-state, spectral and phase-space workflows for closed and open quantum systems. It also includes small command-line helpers that validate model JSON, plan solver settings, sweep cutoffs and audit saved result JSON.
- Units: QuTiP normally sets hbar = 1; Hamiltonian entries are angular frequencies and rates have reciprocal-time units.
- Subsystem order: tensor(A, B, C) fixes indices 0, 1, 2 and the order is preserved throughout.
- Solver choice: for example sesolve for closed unitary dynamics, mesolve for Lindblad open systems, mcsolve for quantum jumps.
- Convergence: sweeps over Fock dimension, time grid, ODE tolerances and trajectory counts where applicable.
Install the skill (ZIP download or Claude apps upload)
Download the ZIP for this skill from this page and unzip it. For a personal install, place the unzipped skill folder in ~/.claude/skills/. For a project-only install, place it in the project's .claude/skills/ folder.
If you use the Claude apps instead, upload the skill under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/
- Project install: .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills
Set up the Python environment
The skill requires Python 3.11 or newer and uses uv, which the skill recommends for environment setup. It pins qutip==5.3.1. The bundled planners and script help run with the Python standard library, but executable simulations need QuTiP installed.
The skill's own setup snippet creates a dedicated environment and installs the pinned package. Plotting needs the graphics extra.
- Python 3.11 or newer
- uv (recommended)
- qutip==5.3.1 (required for executable simulations)
- qutip[graphics]==5.3.1 (optional, for plots)
- qutip-qip, qutip-qtrl and qutip-jax are optional and separately versioned; the skill notes some are pre-release
uv venv --python 3.11
uv pip install "qutip==5.3.1"
uv pip install "qutip[graphics]==5.3.1"Trigger the skill with natural requests
You do not need special syntax. Describe a quantum-system simulation and state what must be explicit, such as physical assumptions, dimensions or convergence. The skill is written for local quantum-dynamics work, so mention that you want a local simulation.
Good requests name the physics: a two-level atom decaying under a Lindblad channel, a Bloch-Redfield model of a weakly coupled bath, or a validation of a saved result file.
- "Simulate a qubit with amplitude damping using QuTiP and report the units and convergence checks."
- "Check this QuTiP model JSON for dimension and tensor-order errors before solving."
- "Audit my saved QuTiP result JSON for trace and positivity problems."
- "Plan solver settings for a stiff open-system model and sweep the Fock cutoff."
Examples of what the agent produces
A typical deterministic open-system run uses mesolve with a collapse operator sqrt(gamma) * sigma_minus. The skill's reference example checks that the excited-state population follows exp(-gamma t) within a stated tolerance and reads result.stats.
For time-dependent drives, the skill recommends trusted Python callables or numeric arrays rather than coefficient strings built from user input, and it advises bounding max_step for narrow pulses.
from qutip import basis, mesolve, sigmam, sigmaz
import numpy as np
excited = basis(2, 0)
result = mesolve(0.5 * 2.0 * sigmaz(), excited, np.linspace(0.0, 20.0, 401),
c_ops=[np.sqrt(0.15) * sigmam()],
e_ops={"excited": excited.proj()})Safety notes: scripts and network
The skill runs scripts, so read them before use if you are working with sensitive environments. The skill states it uses no network service and no credentials. The helper scripts are bounded in size.
The skill is not a hardware execution SDK and cannot control real quantum devices. Its own notes say GPU/MPI, large HEOM hierarchies and experimentally realistic physics are unvalidated, so do not use its predictions for safety-critical or experimental decisions without independent validation.
- Runs scripts: yes
- Uses network: no
- Not for controlling quantum hardware
- Treat GPU/MPI, large HEOM and realistic-physics outputs as unvalidated
Tips for reliable results
Ask the agent to write down units, subsystem order and approximations before any solve. Ask for a convergence sweep over every artificial cutoff, and compare integrators only after rerunning tolerance and invariant checks.
Use the qobj dimensions and shape together: two matrices with the same shape can encode different tensor factorizations.
- Keep the QuTiP version pinned to 5.3.1 for reproducibility.
- Record random seeds for trajectory methods.
- Check ptrace indices: obj.ptrace([0, 2]) keeps those subsystems rather than tracing them out.
- Use the reference notes in the skill folder for composite models and advanced methods.
What people use it for
- Simulating a two-level atom with spontaneous decay under a Lindblad master equation.
- Checking that a composite model's tensor order and dimensions are consistent before solving.
- Running a Bloch-Redfield model of a weakly coupled bath with stated Born-Markov assumptions.
- Auditing a saved result JSON for trace error and negative eigenvalues.
- Sweeping Fock cutoffs and ODE tolerances to confirm a result is numerically converged.
Frequently asked questions
Is the QuTiP agent skill free?
The skill is published under the MIT licence. QuTiP itself is a Python library you install with uv or pip; the source does not state prices for either.
What do I need to run the qutip skill?
Python 3.11 or newer, uv (recommended), and qutip==5.3.1 for executable simulations. Plotting needs the qutip[graphics] extra. The planners and script help use the Python standard library.
Does the skill use the internet?
No. The skill states that no network service or credentials are used. It does run Python scripts, so review them before use.
Can I use this skill to control a real quantum computer?
No. The skill states that QuTiP is not a hardware execution SDK. Circuit functionality lives in separate qutip-qip package.
Where do I install the skill?
Unzip the downloaded folder into ~/.claude/skills/ for personal use or a project's .claude/skills/ folder, or upload it in the Claude apps under Settings → Capabilities → Skills.
Why does the agent ask for units and convergence details?
The skill requires them before solving, because QuTiP does not attach physical units and cannot check that your rates or dimensions are physically consistent. Recording them makes results reproducible.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
This skill suits physics students and researchers running open-quantum-system or Lindblad simulations, and developers who want dimension and convergence checks on QuTiP models. It is not suited to controlling real quantum hardware or to safety-critical predictions, which the skill itself flags as unvalidated. If you only need a quick sketch without documented assumptions, a plain QuTiP script may be enough.
Download the skill