Decision Context: Persistent Behavioral Learning for AI Agents

LLM agents do not learn from corrections across sessions. A user fixes the same mistake at the start of every conversation. Existing agent memory systems handle facts (preferences, entities, conversational history) but not procedural knowledge. We present Forge Decision Context (forge_dc), a system that captures behavioral signals from human corrections and from self-assessed tool-call failures, compiles them into imperative RULE directives, and injects them at the decision points where they apply. We evaluate forge_dc alongside four predecessor approaches on DC-Bench, a new benchmark of 71 scenarios across 9 behavioral dimensions, built from real agent traces. forge_dc reaches 95% compliance, up from 55% with naive retrieval, while injecting roughly $10\times$ fewer tokens per query. The main finding is the "Only Rules Rule": behavioral signals are followed reliably as imperative RULE directives and unreliably as raw text in retrieved memory. The same correction the agent applied in-session is often ignored when surfaced as descriptive text without an anchor. Compilation closes the gap by turning corrections and trace-derived patterns into proactive instructions that apply before any output exists. We also propose a set of design considerations for behavioral memory systems based on what worked and what failed across our experiments.