claimify_skill

Overview

This skill extracts and structures claims from conversations, documents, or debates and turns them into analyzable argument maps. It reveals explicit and implicit assertions, logical relationships (supports/opposes/assumes/contradicts), evidence chains, and tension points so you can evaluate and improve reasoning.

How this skill works

Claimify ingests source text, isolates atomic claims (one testable assertion per claim), and classifies each by type (factual, normative, causal, definitional, predictive, assumption). It then builds a relationship graph linking claims (supports, opposes, assumes, refines, contradicts) and produces outputs in table, graph, narrative, or JSON formats. Optional depth levels control whether the analysis is surface-level, standard, or deep with red-teaming and hidden-assumption mining.

When to use it

• Analyzing a debate or policy paper to surface core claims and gaps
• Red-teaming reasoning or checking for contradictions across a thread or transcript
• Synthesizing meeting notes into a clear argument structure and actionables
• Preparing rebuttals or briefings by mapping supporting and opposing claims
• Transforming unstructured conversation into JSON or graph for downstream tools

Best practices

• Extract atomic claims: one clear assertion per claim
• Be charitable first: steelman arguments before critique
• Label claim types to clarify role in reasoning (factual, normative, causal, etc.)
• Document evidence or source for each claim and flag uncertainty
• Explicitly note assumptions and missing premises that enable claims
• Use appropriate analysis depth: start standard, go deep for strategic red-teaming

Example use cases

• Turn a debate transcript into a claim graph to find contradictions and leverage points
• Convert stakeholder meeting notes into a prioritized claim table with evidence and assumptions
• Evaluate a policy memo: map causal chains and surface missing premises before recommending changes
• Create JSON outputs for programmatic argument analysis or visualization tools
• Run a red-team pass on a product launch argument to expose weak links and hidden assumptions