Skill
unisound-similar-case-retrieval
医生端临床科研 — 相似病例语义检索与可解释排序，锚点病例对候选池做类比推理辅助。
When to use unisound-similar-case-retrieval
Choose if
You have an anchor case summary plus a candidate pool (each with id + summary) and want a top-k ranked list with Markdown-format analogy reasoning. Pick this for research, teaching, or methodological-discussion contexts where the agent's user wants explainable similarity over a black-box embedding search. The skill takes an optional `task_hint` (e.g., "prognosis stratification") to focus the comparison.
Avoid if
You expect direct hospital-EHR connectivity — the README explicitly states the skill "does not interface with real hospital records"; candidates must be pre-prepared and de-identified. Also avoid as clinical advice — the README scopes output to "research, teaching, or methodological discussion reference only, not clinical advice" and requires ethics approval before using real patient data.
Risk Flags
- MEDIUM auth Requires an internal `appkey` (Bearer token) for the Unisound maas-api.hivoice.cn medical LLM endpoint (model u1-insuremed).
- MEDIUM scope README explicitly states the skill does not interface with real hospital records — candidate cases must be pre-prepared and de-identified by the caller.
- MEDIUM scope README explicitly states output is for research, teaching, or methodological discussion only and is not clinical advice; real-patient use requires deidentification and ethics approval.
Cost
Type: Unknown
Distribution
- ClawHub
unisound-similar-case-retrieval- License
- MIT-0