Design verification strategies by diagnosing observability and mapping oracle types to loop tiers.
npx skills add https://github.com/hashintel/brunch --skill ln-oracles
Please help me install this Agent Skill. Skill: ln-oracles Source: https://github.com/hashintel/brunch/tree/main/.agents/skills/ln-oracles Command: npx skills add https://github.com/hashintel/brunch --skill ln-oracles
It helps design effective verification strategies by diagnosing observability, selecting appropriate oracle families, and mapping them to verification loop tiers, ensuring system correctness.
Review your system's verification goals, assess observability and reproducibility, then select oracles and assign them to loops accordingly.
High-intent search queries and answers about installing and using this skill.
To design a verification strategy for reliable AI systems, assess observability and reproducibility, extract structural and behavioral claims, then select appropriate oracle families and map them to verification loop tiers.
An oracle in AI system correctness testing is a mechanism used to validate claims. Multiple oracle types are selected and integrated to systematically enhance reliability and ensure comprehensive verification coverage.
To identify blind spots and verification gaps, diagnose observability, reproducibility, and controllability. This diagnostic assessment highlights risks and maps verification coverage to match project complexity.
Map oracle types to verification loop tiers by first scoring observability and extracting claims, then choosing suitable oracle families and assigning them to specific loop tiers based on project risk and slice requirements.
The best way to extract verification claims is to identify what needs to be proved across structural, behavioral, and compositional levels, ensuring verification coverage aligns with system complexity.
Yes, you can use this diagnostic framework for slice verification design. It specifies verification approaches for individual project slices to ensure targeted coverage and mitigate identified risks.
CAPHTECHDefine a minimal set of diagnostic and verification practices covering six failure modes.
dsteven12Coordinate parallel subagents to synthesize unified recommendations with tradeoffs.
daidrEnrich event logs with environment data, feature toggles, or tenant identifiers.
rajyeole6Audit smart contracts for reentrancy, oracle manipulation, and access control vulnerabilities.
