lalala5678

Two skill sets are offered: memory-systems covers implementing agent memory, cross-session state persistence, entity tracking, temporal knowledge graphs, and vector stores; advanced-evaluation covers LLM-as-judge setup, output comparison, evaluation rubrics, and bias mitigation in assessment pipelines.

Engineers and researchers building conversational agents that need persistent memory, plus ML practitioners designing automated quality assessment pipelines. The author is a master's student at BUPT (Beijing University of Posts and Telecommunications) focused on agent memory and model evaluation topics.

Each skill activates from natural-language requests defined in its frontmatter. Memory-systems triggers on phrases like 'implement agent memory' or 'persist state across sessions'; advanced-evaluation triggers on 'implement LLM-as-judge', 'compare model outputs', or 'mitigate evaluation bias'.

For memory-systems, familiarity with memory architecture, entity memory, temporal knowledge graphs, and vector stores is assumed. For advanced-evaluation, you should understand direct scoring, pairwise comparison, position bias, and evaluation pipeline design.