使用方法
- 复制安装命令，在自己的智能体环境安装前先检查源码包。
- 从原始文档核对所需工具、凭据和支持的宿主。
- 说明输入材料与期望交付物，并对照来源检查输出。
npx skills add https://github.com/yogsoth-ai/de-anthropocentric-research-engine --skill detect-breakpoint开始前需了解
- 安装不包含模型订阅、API 额度或访问文件的权限。
- 仓库星数和安装量反映活跃情况，不等于独立质量或安全测试。
仓库原始内容
按源仓库内容呈现，保留标题、案例、代码、表格、链接以及原文引用的演示图片。
detect-breakpoint
Purpose
Locate the point or region where performance, validity, or a claim changes regime.
Input contract
yaml
required: [ordered_observations, target_metric, flip_or_regime_rule]
optional: [candidate_axis, smoothing_rule, uncertainty_bounds]
constraints: [ordering and metric direction are explicit]Procedure
- Sort observations along the declared axis and validate comparability.
- Detect abrupt flips, slope changes, saturation, or regime transitions.
- Estimate boundary region and attach uncertainty.
- Return witness observations and follow-up tests.
Output contract
yaml
produces: [breakpoint_region, regime_label, witness_observations, uncertainty, follow_up_tests]
delta_fields: [findings, uncertainties, open_questions]Quality gates
- A breakpoint requires observations on both sides or an explicit one-sided limitation.
- Metric direction and flip rule are declared before detection.
- Smoothing or interpolation never replaces raw witnesses.
Parameterization
Caller supplies observation schema, axis ordering, metric, flip/regime rule, smoothing, and uncertainty policy.
Failure and counterexamples
Reject boundaries inferred from unvalidated ordering, single noisy points, or hidden smoothing.
Provenance map
- concept: stress-test/breakpoint-detection
- concept: deep-insight/controlled-perturbation
