microbiomedata/nmdc-metadata-suggestor-ai-tool

build-study-context

Use this skill to assemble an evidence bundle (context texts) from a parsed submission or ETL study context, including DOI abstracts and PDF content.

supplement-retrieval

Use this skill to retrieve supplementary materials for a manuscript DOI, prioritizing the file types useful for characterizing NMDC submissions and samples (tabular metadata and supplementary documents). Performance-first, single attempt, no repeated retries.

doi-ingestion

Use this skill to fetch abstract or description text for a DOI using a source waterfall (datacite, crossref, openalex, pubmed, and repository-specific sources like osti, ess-dive, jgi).

metadata-suggestion-pipeline

Use this skill to suggest metadata field values for a full NMDC submission object by parsing submission data, fetching DOI/PDF context, loading schema, and returning cited LLMOutput JSON.

nmdc-metadata-suggestor

Use this skill to orchestrate NMDC metadata suggestions — detects input shape, routes to env-triad and/or metadata-suggestion-pipeline, and returns a merged LLMOutput JSON. Entry point for all metadata suggestion requests.

env-triad

Use this skill to suggest env_broad_scale, env_local_scale, and env_medium values for NMDC biosample records using study context, DOI abstracts, and schema enumerations.

submission-parser

Use this skill to extract structured fields (DOIs, study name, description, protocol info, MIxS extensions) from a raw NMDC submission portal object. Called at the start of metadata-suggestion-pipeline and env-triad when a submission_object is present.

pdf-ingestion

Use this skill to retrieve PDF content for publications via web_fetch (agentic path).

schema-context

Use this skill to retrieve and format NMDC submission schema context (MIxS interface slots and env triad values) for use in LLM recommendation prompts.