person-context manages real-world individual relationships in depth — the people that matter most to the user (business partners, key clients, friends, family, and the user themselves). It treats observation as cognitive traits, builds a per-person markdown dictionary, and auto-injects the relevant dictionary into Claude's context whenever the person comes up in a prompt. Unlike persona-switching plugins designed for market research or role-play, person-context targets continuous, longitudinal management of a small number of important real persons. Three components: 1. Dictionary templates — a 9-chapter catalog plus custom chapters per person. 2. Inject hooks — UserPromptSubmit auto-injection so the dictionary is loaded before Claude responds; PreToolUse verification that physically blocks responses where the dictionary was not actually read. 3. Critical Feedback Patterns — four response rules that co-fire with the dictionary (careful interpretation, dictionary accuracy, pressure-test, obvious-judgment). Key differentiation vs existing claude-personas / claude-persona / TinyTroupe: real persons + the user (not synthetic personas), an observation-driven update loop (extract -> analyze -> apply), and full Japanese + English documentation. v0.1.0 released 2026-05-15; v0.1.1 (2026-05-17) adds full English README, template, three sample dictionaries, and CHANGELOG.