Basal Ganglia Memory

**Quick judgement** This skill is an experimental, under-development module that adds basal-ganglia–style habit and procedural memory atop standard agent memory. It is promising for researchers and tinkerers building more lifelike, long-lived agents, but it is not yet mature enough to depend on for production-critical behavior. **What it’s good for** - Prototyping agents that learn routines and preferences from repetition and reward signals. - Building richer cognitive stacks alongside other AI Brain skills (e.g., hippocampus, amygdala). - Research on habit formation, long-term behavioral drift, and reinforcement effects in agents. **Key risks / caveats** - **Under development:** APIs, data schemas, and behavior may change, and documentation may be incomplete or unstable. - **Behavioral lock-in:** Poorly tuned rewards or signals can cause agents to over-commit to suboptimal habits (“it worked once, so I always do it”). - **Debugging difficulty:** Learned procedural patterns may be opaque, making it harder to understand why the agent “just does it this way.” - **Safety / alignment nuance:** Habit- and reward-driven modules can amplify reward-hacking or undesirable shortcuts if not carefully monitored and constrained. **Recommended scenarios** Use this skill when you are experimenting with long-lived or autonomous agents and want them to develop stable routines and preferences over time, and you are comfortable dealing with experimental, evolving code. Avoid relying on it as the sole controller for safety-critical or high-stakes applications until the implementation and documentation are more mature and thoroughly tested.