reflexion-memorize

To consolidate agent reflections into a structured knowledge base, extract key insights from critiques and categorize them by impact into an evolving playbook like AGENTS.md. This prevents context collapse and ensures actionable knowledge accumulation.

Agentic context engineering for memory consolidation is the process of transforming raw insights from conversations and reflections into durable, verifiable knowledge. It continuously refines a playbook of best practices to improve future agent performance.

To prevent context collapse when accumulating agent knowledge, apply strict curation rules that require extracting key insights and categorizing them by impact. This ensures the evolving playbook remains actionable and verifiable.

The best way to curate insights from problem-solving sessions into a playbook is to distill key lessons learned from agent interactions and critiques into a structured file like AGENTS.md, ensuring similar future challenges are handled more effectively.

You need to update your agent's playbook with reflection insights after completing a complex problem-solving session involving several agent interactions and critiques. This continuously refines best practices, patterns, and anti-patterns for future tasks.

No dependencies are required to curate an evolving playbook. The memory consolidation process operates independently to extract key insights from reflections and structure them into actionable knowledge within AGENTS.md.