The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill includes functionality for reading and processing external molecular structure files, which creates a potential surface for indirect prompt injection if the files are sourced from untrusted locations.
Ingestion points: The function qchem.read_structure('molecule.xyz') in references/quantum_chemistry.md allows the agent to ingest external file data.
Boundary markers: No specific delimiters or safety instructions are provided to the agent for handling the content of these external files.
Capability inventory: The skill utilizes local computational resources for simulations and network-connected quantum hardware backends via official plugins for IBM Quantum, AWS Braket, and Google Cirq.
Sanitization: There is no evidence of explicit sanitization or validation of the ingested file contents before they are used in quantum simulations.
[DYNAMIC_EXECUTION]: The documentation demonstrates the use of Just-In-Time (JIT) compilation through the Catalyst library to optimize quantum circuit execution performance.
Evidence: Examples in references/advanced_features.md and references/devices_backends.md show the use of the @qjit decorator and the catalyst package to compile functions into optimized executables.
[EXTERNAL_DOWNLOADS]: The instructions guide the installation of various quantum computing plugins and libraries from public registries.
Evidence: The skill recommends using uv pip install for standard packages such as pennylane-qiskit, amazon-braket-pennylane-plugin, and pennylane-azure. These are official plugins from well-known technology providers and trusted organizations.

pennylane