opentargets-skill

Send a JSON object with a `query` field (or `query_path` for long documents) to the opentargets_graphql.py script via stdin. It posts to the Open Targets v4 GraphQL endpoint and returns a compact summary with configurable `max_items` and `max_depth` truncation.

Run opentargets_disease_heatmap.py with an `ensembl_id`, optional `page_size`, `max_pages`, and `disease_name_filter`. It paginates `associatedDiseases`, collects `datasourceScores`, and returns a matrix of diseases by datasource with labeled columns.

The scripts require only the `requests` package beyond the Python standard library. If requests is missing, the scripts return a `missing_dependency` error code instead of failing silently.

Yes, set `save_raw` to true and optionally provide `raw_output_path` in the input JSON. The full GraphQL response is written to that path, defaulting to /tmp/opentargets-graphql.json.

The heatmap reflects evidence-source breadth from `datasourceScores`, not causal proof of target-disease association, mechanism, or direction of effect. It also paginates up to `max_pages`, so very large result sets may be truncated with a warning.