What problem does it solve? Conversation data with reply chains hides social structure that is hard to see manually: who talks to whom, whether interactions are mutual, and whether a user changes their writing style depending on the audience. This Skill turns raw reply chains into a measurable directed graph so those patterns can be quantified. ## Core Features & Use Cases - Directed Graph Construction: Extracts reply pairs from conversation data and builds a weighted NetworkX DiGraph with filtering for trivial interactions. - Ego Network & Reciprocity Analysis: Extracts a target user's ego network, classifies top interlocutors (balanced, ego-dominant, one-directional), and measures reciprocity at global, ego, and per-alter levels. - Audience-Dependent Voice Shift Detection: Compares linguistic features (word length, formality, vocabulary richness) across interlocutors to detect accommodation patterns, plus community detection for larger graphs. - Use Case: Given a Reddit comment export, map a user's interaction network, identify their five most frequent conversation partners, and determine whether their tone shifts when replying to different people. ## Quick Start Use the network-social-graph skill to analyze the reply chains in my conversation export and write the findings report.
