logic-hunter_skill

Overview

This skill performs hardcore logical verification and provenance tracing using a "Golden Triangle" knowledge-mining framework. It hunts for truth by prioritizing first-hand sources, scoring evidence, and flagging untraceable conclusions as hypotheses. The output is a one-page, slide-style report highlighting confidence, evidence weights, attack vectors, and entropy-driven risks.

How this skill works

It parses user queries to extract core variables and remove semantic noise, then retrieves primary sources via web search and research tools. Retrieved items are categorized (primary/secondary/tertiary) and fed to a scoring engine that computes a confidence metric using reliability, support, and entropy. Finally a red-team simulation probes survivorship bias, causality reversal, and other weak points before producing a concise report.

When to use it

• Verify the credibility of a specific claim or prediction
• Assess how much evidence supports a controversial statement
• Trace conclusions back to primary sources for reporting or due diligence
• Produce a compact slide-style credibility summary for meetings or briefs
• Run a red-team check on internal analyses or public narratives

Best practices

• Provide a clear, testable claim or question to focus retrieval and scoring
• Prioritize requests that seek source provenance and confidence, not just summary
• Supply known documents or links to accelerate primary-source classification
• Treat outputs as structured assessments, not absolute truth—investigate flagged hypotheses
• Request follow-up deep dives when primary evidence is scarce or entropy is high

Example use cases

• Checking whether a market claim is backed by company filings and peer-reviewed studies
• Evaluating the credibility of a viral social-media assertion before publication
• Assessing confidence in policy or regulatory claims citing mixed media sources
• Summarizing evidence weight and attack vectors for leadership briefings
• Red-team testing a research conclusion to surface hidden bias or causal errors