Cross-check claims against multiple primary sources and generate audit files.
npx skills add https://github.com/KodingDev/meridian --skill triangulate-kodingdev
Please help me install this Agent Skill. Skill: triangulate Source: https://github.com/KodingDev/meridian/tree/main/skills/triangulate Command: npx skills add https://github.com/KodingDev/meridian --skill triangulate-kodingdev
Triangulate enforces rigorous truth-telling by cross-checking specific-value or behavior claims against multiple primary sources before any assertion or action.
Audit a claim against multiple primary sources before asserting or acting on it.
High-intent search queries and answers about installing and using this skill.
To verify API behavior claims, you cross-check specific values against multiple primary sources before asserting them. This requires two sources of different kinds for high confidence, outputting inline audit rows and a full audit file to document all evidence and disagreements.
Single-source verification fails because it lacks the confidence-gating needed to prevent risky decisions. Triangulation requires two sources of different kinds to surface doubt and ensure evidence-backed outputs, whereas a single source is insufficient for high confidence.
The best way to audit design tokens is performing tiered grounding: executing quick Tier-1 checks or escalating to heavier Tier-2 audits. This generates inline audit rows and a full audit file that documents sources, confidence levels, and potential disagreements.
No dependencies are required to cross-check config fields. The process requires two sources of different kinds for high confidence, but these are primary reference sources rather than external software dependencies or library integrations.
You should not use claim verification when a single source is sufficient or when quick action is needed without evidence documentation. Use it only for specific-value or behavior claims where high confidence, source documentation, and disagreement tracking are required.
Yes, you can generate a ground truth audit file for config fields. The process produces inline audit rows and writes a full audit file to document the evidence, source confidence levels, and any potential disagreements found during the cross-check.
ShubhamsabooGuide structured fact-checking through claim identification, evidence gathering, and credibility assessment.
openaiNormalize extracted financial rows into a model-ready long-form schema with QA artifacts.
MykhailoDmytriakhaVerify factual claims with primary sources, confidence scores, and Loop150 escalation.
stephendorConvert verified research sections into web pages, talks, and posters.
