The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill describes functions that retrieve data from external scientific sources, such as PubChem (via derive_annotation_from_compound_name) and Universal Spectrum Identifier (USI) repositories (via load_from_usi). These are legitimate features for metadata enrichment in metabolomics research.
[REMOTE_CODE_EXECUTION]: Matchms supports data serialization via the Python pickle module (load_from_pickle and save_as_pickle). While this is a standard method for preserving the state of complex Python objects like mass spectra, users should ensure they only load files from trusted sources to avoid potential risks associated with untrusted data deserialization.
[PROMPT_INJECTION]: The skill processes external data files (mzML, MGF, MSP, JSON) which constitutes a vulnerability surface for indirect prompt injection if the metadata contains malicious instructions. However, the skill provides extensive filtering and harmonization functions (e.g., default_filters, clean_compound_name) that help standardize and sanitize the input data.
Ingestion points: Data is ingested from external files via functions like load_from_mgf and load_from_json (found in references/importing_exporting.md).
Boundary markers: No explicit prompt boundary markers or instructions to ignore embedded commands are present in the provided examples.
Capability inventory: The skill can write files (save_as_mgf) and perform network requests for metadata enrichment (PubChem, USI).
Sanitization: The library includes robust filtering mechanisms such as clean_compound_name and harmonize_undefined_smiles (found in references/filtering.md) which process and validate metadata.

matchms