What problem does it solve? LLMs fail at temporal reasoning over long organizational histories: attention degrades across hundreds of chronologically ordered events, and events retrieved without causal context poison the model's conclusions. This Skill provides the architecture patterns and query workflows to give agents structured temporal storage and curated causal retrieval instead of raw timeline dumps. ## Core Features & Use Cases - Three Temporal Query Types: Sequence queries (what happened between A and B), causal queries (what caused X), and counterfactual queries (what if decision D had been different), each with a defined retrieval strategy. - Event Graph Architecture: Store institutional events as nodes linked by CAUSED, TRIGGERED, PRECEDED, and SUPERSEDED_BY edges, with Cypher traversal patterns for Neo4j-style graph databases. - Windowed Context Synthesis: Compress distant history into summaries while keeping recent events in full detail, so long-horizon queries fit the context window. - Use Case: An agent is asked "what decisions led to us adopting OAuth2?" It locates the adoption event, traverses causal predecessors up to 4 hops, assembles a structured causal chain, and returns a plain-language narrative of the decision history. ## Quick Start Use the engineering-temporal-reasoning skill to design a causal event graph and answer what decisions led to our current authentication architecture.
