continuous-learning-v2

Session observations are converted into reusable instincts by applying hooks that identify recurring patterns and evaluate them with confidence scoring for local persistence.

An instinct is an atomic observation extracted from a session and scored for confidence, serving as the foundational primitive that evolves into skills, commands, or agents.

Extracted instincts evolve into commands or agents through an evolution pipeline that clusters related instincts and organizes them into higher-level primitives like skills, commands, and agents.

Yes, an observer workflow is required to process session observations, as the hooks rely on this active observer to identify patterns and persist them locally for evolution.

Yes, YAML-frontmatter skill metadata is required to define the structure needed for the evolution pipeline to successfully transition extracted instincts into organized skills and commands.

Local persistence limits instinct sharing to a single environment, meaning learned patterns and evolved primitives are not automatically synced or distributed across different machines or team members.