instrument-data-to-allotrope

instrument-data-to-allotrope is a single skill from Anthropic's official open-source knowledge-work-plugins repository. It belongs to the bio-research plugin and helps scientists, laboratory teams, and data engineers turn heterogeneous laboratory instrument exports into a standardized representation. The official source is https://github.com/anthropics/knowledge-work-plugins/tree/main/bio-research/skills/instrument-data-to-allotrope. The repository presents the skill as a Markdown-based set of working instructions; it is not an independent server and it is not a complete LIMS integration.

Purpose and workflow

According to the provider, the workflow starts by detecting an instrument type from the file contents or by using a type selected by the user. Supported input families can include PDF, CSV, Excel, and TXT. The process then tries native parsing with the allotropy library first. When no suitable native parser is available, a flexible fallback parser can infer columns, units, and metadata from the document structure. PDF input is prepared through table extraction. This order matters because a native parser generally preserves more instrument-specific meaning than a generic interpretation of a table.

Outputs for research and data platforms

The main target is the Allotrope Simple Model, or ASM, as JSON with semantic structure and ontology references. According to the provider, this output is suited to LIMS systems, data lakes, and long-term archival. The skill can also produce a flattened two-dimensional CSV where each measurement occupies a row and shared metadata is repeated. That representation is useful for quick analysis, spreadsheets, and systems that do not accept JSON. A third handoff can be exportable Python code for data engineers to inspect, adapt, and evaluate in notebooks or production pipelines.

Raw and calculated data

A central quality concern is separating direct instrument measurements from derived results. The skill places raw measurements in a measurement-document and calculated values in a calculated-data-aggregate-document. Derived values should retain provenance through a data-source-aggregate-document structure. This lets a concentration calculated from absorbance or a viability value derived from cell counts point back to its source measurements. The distinction supports reproducibility, but it must not be mistaken for formal approval or validation by an accredited laboratory.

Instruments and validation

The official description names cell counters such as Vi-CELL BLU, Vi-CELL XR, and NucleoCounter, spectrophotometers such as NanoDrop and Lunatic, plate readers, ELISA systems, qPCR instruments, and chromatography platforms. Actual coverage depends on the installed allotropy parser and the export format. The skill recommends checking available vendors first and consulting a reference ASM document when needed. Before handoff, the generated JSON should be checked with the intended validator. According to the provider, the validation rules are based on the ASM specification and examine technique selection, field naming, traceability, unique identifiers, units, and required metadata. Warnings for unknown techniques or units can be permitted intentionally, but they require human assessment.

Boundaries, safety, and fit

The provider describes this as an example skill that demonstrates schema transformations and data engineering patterns. It does not replace instrument qualification, scientific data review, or testing of the target system. Proprietary export variants, damaged tables, missing units, or ambiguous field classification can produce incomplete results. When a field mapping is unclear, the skill asks the user to clarify its meaning rather than inventing plausible semantics. Laboratory files may contain personal, clinical, or commercially sensitive information. Before transmission, review approved clients, storage locations, access controls, retention, and the possible model or service provider. A local input file does not automatically mean that a connected Claude client or model will keep all information local. The repository is licensed under Apache-2.0. This description reflects the source checked on September 9, 2026 and is not evidence of regulatory suitability.