Deep research — BRCA1 DNA repair literature survey (English)
PASS · rubric SATISFIED545sprompt: enartifact: en
Prompt (verbatim)
no-interview light-research: Using the bioresearcher-deep-research skill, survey the recent article landscape on BRCA1 DNA repair. Keep it to at most 2 research aspects and at most 10 biomcp tool calls per aspect worker. Cite sources with PMIDs.
What happened
- Loaded the deep-research skill and smoke-tested the MCP connection (biomcp_gene_search) before any research.
- Fanned out two parallel Task workers (BRCA1 HR mechanisms + BRCA1 PARPi resistance); each appended every citable source to a per-aspect evidence ledger and passed the check --markers completion gate.
- Orchestrator merged + NCBI-verified the ledgers (sources.jsonl: 34 records, 33/33 markers resolved), drafted with [@pmid:...] cite-key markers; render numbered all 30 citations and generated the Vancouver References with citation locators (volume/issue/pages) wherever the source provides them; vet-references structural audit PASS; interactive HTML rendered by default.
Evidence
Objective checks
checks passed: 7/7 · one rubric flag adjudicated SATISFIED post-run (recorded in result.json)
| Fan-out mode | Parallel fan-out via the harness Task tool (2 subagent workers, captured live from the opencode session DB) |
|---|---|
| Provenance | d818c85 · opencode 1.18.30 · biomcp@1.4.0 |
Replay
node demos/run-demo.mjs --only agent-deep-research-en --publish
Manual-run only — this replay spends LLM tokens (real opencode CLI + network).
In the repository: README · transcript · result.json · artifact directory (permalink)