Agent Memory Requirements Directory

Type: kb/types/index.md

← Parent

• Activate Behavior-Changing Memory Before The Mistake (note) - Behavior-changing memory must activate before relevant actions rather than waiting for explicit retrospective search
• Create Memory Directly (note) - Direct memory creation preserves live understanding by writing useful artifacts before later trace extraction loses structure
• Evaluate Memory By Effects, Not By Existence (note) - Memory should be evaluated by downstream effects on tasks, artifacts, answers, behavior, context efficiency, and source alignment
• Import External Knowledge Into Internal Form (note) - Agent memory systems need import paths when authoritative project knowledge already exists outside the memory substrate
• Keep Memory Roles And Compiled Views From Drifting (note) - Generated cues, prompt files, indexes, and assistant-specific views need source-of-truth rules so they do not drift into authority
• Make Authority Explicit (note) - Memory architecture must state who can read, write, promote, activate, enforce, revise, and retire memory across risk levels
• Preserve Evidence Without Making History The Next Context (note) - Trace retention should preserve evidence for audit and extraction without making raw history the agent's default context
• Promote Only When Future Value Exceeds Maintenance Cost (note) - Candidate memory should become durable only when future retrieval or activation value exceeds review and maintenance cost
• Retire, Redact, Supersede, And Relax Memory (note) - Memory systems need lifecycle operations for redaction, decay, supersession, retirement, relaxation, and temporal validity
• Serve Multiple Consumers, Not One Retrieval Interface (note) - Memory systems need multiple surfaces because acting, scheduling, review, learning, governance, and active work consume memory differently
• The adaptation survey corroborates memory requirements but misses artifact-role governance (note) - The agentic-adaptation survey supports the memory requirements map by treating memory and skills as adaptive tools, but it needs artifact-role governance to become design guidance
• Use Trace-Derived Extraction As Meta-Learning (note) - Trace-derived extraction is an after-the-fact learning path that must respect signal quality, review, and artifact versus weight-learning boundaries