qml-evaluation

Quantum ML Skills

Reusable QML skills with PennyLane-first design, PyTorch-first training, and exporter-based compatibility for OpenCode and Claude Code.

These skills are maintained from a single source library so they can support OpenCode today and Claude Code through a compatible export layer without splitting the docs or source structure.

Installation

OpenCode

Install into a local project:

bash install/install_opencode.sh --project-root .

Install globally:

bash install/install_opencode.sh --global

Global installs go into:

~/.config/opencode/skills/qml-skills/

This keeps the skills namespaced and avoids resetting the whole global OpenCode skills root.

If you prefer not to call the shell installer directly, see:

• examples/use-with-opencode.md

You can validate an install with:

python install/doctor_opencode.py --path .opencode/skills

Claude Code

Claude Code support is provided through:

• root CLAUDE.md
• project .claude/settings.json
• a GitHub-hosted marketplace view at repo root under .claude-plugin/ and plugins/
• generated exports under skills/qml/exports/claude-code/
• a local plugin-style marketplace under skills/qml/exports/claude-marketplace/

To generate Claude-compatible output:

python skills/qml/exporters/export_claude_code.py

To generate a local Claude marketplace:

bash install/install_claude_marketplace.sh --project-root .

To sync the GitHub-hosted marketplace view into the repository root:

python skills/qml/exporters/export_claude_marketplace.py --sync-hosted-root

Then add it inside Claude Code:

/plugin marketplace add ./.claude/marketplaces/qml-skills

To add this repository from GitHub after the root marketplace view is published:

/plugin marketplace add TQuang122/quantum-ml-skills

For faster Git checkout of the hosted marketplace only:

/plugin marketplace add TQuang122/quantum-ml-skills --sparse .claude-plugin plugins

Recommended install order:

/plugin install qml-common@qml-skills
/plugin install qml-core@qml-skills
/plugin install qml-backends@qml-skills
/plugin install qml-evaluation@qml-skills
/plugin install qml-research@qml-skills

Install plugin bundles sequentially. Parallel install attempts can race on local Claude plugin state.

The canonical source of truth remains:

skills/qml/

Skills

Skill	Description
qml-foundations	Frame QML problems before implementation.
qml-pytorch-router	Route ambiguous PennyLane + PyTorch QML requests to the correct implementation skill.
pennylane-qnn	Build and refactor PennyLane-first hybrid quantum models.
qml-pytorch-interface	Clean PyTorch tensor, parameter, and prediction boundaries around PennyLane models.
qml-pytorch-training	Build reusable PennyLane + PyTorch training workflows.
qml-pytorch-performance-patterns	Improve performance for PyTorch-based QML workloads.
pennylane-qiskit-backends	Add Qiskit-backed execution while keeping PennyLane as the authoring layer.
qiskit-machine-learning-interop	Explore native Qiskit Machine Learning abstractions when plugin-backed execution is not enough.
qml-cross-framework-benchmarking	Compare QML branches and backends fairly.

Advanced docs

• skills/qml/README.md
• skills/qml/REQUEST_PATTERNS.md
• skills/qml/ROUTING.md
• skills/qml/STARTER_WORKFLOW.md
• skills/qml/EXPORT_STRATEGY.md
• examples/use-with-opencode.md
• examples/use-with-claude-code.md
• CLAUDE.md
• CONTRIBUTING.md

Release

• CHANGELOG.md
• RELEASE_CHECKLIST.md
• RELEASE_NOTES_v0.1.0.md

License

MIT