Rank hypotheses for root-cause analysis using a four-layer validation workflow.
npx skills add https://github.com/iheCoder/Lib --skill hypothesis-validator
Please help me install this Agent Skill. Skill: hypothesis-validator Source: https://github.com/iheCoder/Lib/tree/main/skill/hypothesis_validator Command: npx skills add https://github.com/iheCoder/Lib --skill hypothesis-validator
Hypothesis Validator helps incident responders systematically prune unreliable explanations by enforcing a four-layer causal validation: boundary, temporal, magnitude, and scope, plus counterfactual checks, to ensure root cause reasoning stays grounded in data.
将证据包和假设输入到 Validator，按时间、量级、范围进行评估，并输出综合分数与扩展建议。
High-intent search queries and answers about installing and using this skill.
Root-cause analysis hypotheses are validated through a four-layer workflow evaluating temporal, magnitude, scope, and counterfactual evidence. This structured approach prunes false hypotheses by scoring causal relationships against available data, ensuring incident reasoning remains grounded.
Counterfactual validation in root-cause analysis tests whether the observed anomaly would still occur if the suspected cause were absent. It serves as one of four structured scoring layers, alongside temporal, magnitude, and scope checks, to prevent overfitting to noisy signals.
Multiple incident hypotheses are ranked by applying structured scoring across temporal, magnitude, scope, and counterfactual dimensions. The validation process outputs a JSON object containing comprehensive scores for each hypothesis alongside a recommended decision.
When evidence is insufficient to confirm a root cause, the validation process suggests which world contexts to query next. It recommends investigating specific operational areas like deploy, scaling, job, or config environments to gather additional supporting data.
Hypothesis validation is designed for incidents with anomalies where multiple hypotheses exist and evidence is available. Applying it to a single suspected cause limits the comparative scoring benefits needed to effectively prune unreliable explanations.
markusstrasserGenerate competing hypotheses and eliminate them to reveal the surviving explanation.
jfmyers9Diagnose bugs, incidents, and CI failures with read-only root-cause reports.
Arry8Generate structured AI operation audit summaries with run metadata, decision traces, and anomaly signals.
YangYuS8Identify and isolate root causes of bugs before proposing fixes.
