protocol-design
Routes biomedical research questions into ethically aware, feasibility-conscious study protocols covering aims, design, endpoints, bias, power, validation, and translational planning.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- protocol-design
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Routes biomedical research questions into ethically aware, feasibility-conscious study protocols covering aims, design, endpoints, bias, power, validation, and translational planning.
- disable-model-invocation
- true
02 资源文件 3
- 参考文档references/design-checklists.md7.9 KB
- 参考文档references/troubleshooting.md7.2 KB
- 参考文档references/workflows.md8.3 KB
03 SKILL.md 正文
Protocol Design
Use this route when a researcher needs to turn a biomedical idea into a study-design plan: aims and hypotheses, design family, eligibility and endpoints, bias/confounding controls, sample-size logic, validation, or a translational/specialized protocol. This is protocol planning, not patient care, statistical execution, or manuscript prose.
Route and boundary
- Frame the question. State the population/model, exposure or intervention, comparator, primary outcome, time horizon, intended estimand, and whether the claim is descriptive, associative, predictive, mechanistic, or causal-leaning. Create one coherent primary aim; subordinate secondary and exploratory aims.
- Select the study design. Choose among experimental, clinical cohort, case-control, diagnostic/prognostic, RWE, translational, omics, or genetic (including MR) designs based on temporal logic, available data/specimens, recruitment, and the claim—not on habit. Define time-zero, eligibility, baseline window, exposure/intervention, follow-up, censoring, and endpoints.
- Stress validity and feasibility. Separate baseline from post-baseline variables; identify selection, confounding, immortal-time, measurement, missingness, multiplicity, transportability, and competing-risk threats. Treat effect sizes, event rates, variance, follow-up, data capture, and validation resources as unknown until supported. Obtain ethics, privacy, safety, recruitment, specimen, site, and operational feasibility review.
- Plan power and validation. Let the primary endpoint/estimand drive formal, range-based, event-driven, precision, or fixed-N feasibility planning. Specify internal versus independent, temporal/site/external, technical/orthogonal, functional, and translational validation tiers; never call resampling external validation or infer causal proof from support alone.
- Produce a bounded handoff. Return assumptions, decisions, alternatives, feasibility blockers, minimum credible protocol, analysis/validation needs, and what remains unverified. Route live evidence retrieval and citation verification to evidence-insight, executable modeling/data checks to data-analysis, and protocol/manuscript/grant prose to academic-writing. Ethics/privacy or operational concerns may also need operations-and-audit.
Read the bundled references
- workflows.md for the end-to-end route, design families, outputs, and specialist handoffs.
- design-checklists.md for aims, eligibility, endpoints, validity, power, validation, translation, ethics, and feasibility gates.
- troubleshooting.md for missing inputs, design mismatch, unsupported assumptions, and recovery actions.
Safety and evidence limits
Do not provide patient-specific diagnosis or treatment, invent literature, registry/database capture, event rates, sample sizes, assay feasibility, validation cohorts, or ethical approval. Mark candidate resources and assumptions explicitly. A protocol draft is not an IRB/REC/IACUC submission, statistical analysis execution, or clinical decision. If a critical design input is missing, ask focused questions or proceed only with a clearly labeled provisional design and an explicit stop condition.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).