matchms

Warn

Audited by Gen Agent Trust Hub on Mar 31, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill and the underlying 'matchms' library explicitly support and encourage the use of Python's 'pickle' format for fast loading and saving of processed spectra (e.g., load_from_pickle, save_as_pickle). Deserializing data using the 'pickle' module is inherently unsafe as it can be exploited to execute arbitrary code. An attacker could provide a malicious .pkl file that, when loaded by the agent, compromises the host environment.
  • [EXTERNAL_DOWNLOADS]: The skill includes functionality to fetch data from external sources during processing. Specifically, derive_annotation_from_compound_name retrieves chemical information from PubChem (NIH), and load_from_usi fetches spectral data from online metabolomics repositories using Universal Spectrum Identifiers. While these target well-known scientific services, they involve automated external network requests triggered by input data.
  • [COMMAND_EXECUTION]: The installation instructions recommend using uv pip install, which involves executing system-level commands to manage dependencies. While standard for library setup, it requires appropriate environment permissions.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 31, 2026, 08:20 AM
Security Audit — agent-trust-hub — matchms