Basal Ganglia Habit Recorder · Behavior Automation and Habit Formation

Problem

In long-running AI agent workflows, many actions are not novel: the same cleanup, lookup, templated generation, or style-dependent output repeats. Re-running full intent decoding, prechecks, and confirmation every time makes the path longer and forces users to repeat the same preferences. habit-loop addresses the repetition cost of high-frequency workflows by turning verified task paths into reusable rules instead of re-deriving them on every turn.

How It Works

The skill works around a local habit store at .workbuddy/dreams/habit-store.json. Its core loop is:

• Automatic formation: when a similar task appears at least three times, or task-frequency-tracker.json reveals a recognizable pattern, the system marks the candidate task as a habit rule.
• Hit execution: later matching tasks can run the verified path directly, skipping some prechecks and confirmation steps, reducing flow overhead in the low single-digit second range.
• Learning and maintenance: user overrides are logged as exceptions; repeated hits strengthen confidence; stale rules decay so outdated logic does not pollute later execution.

It can also coordinate with related skills: high-confidence, high-frequency habits may be suggested for promotion to formal skills, while working memory, REM insights, and style preferences can seed habit candidates.

Boundaries

This mechanism fits repetitive, well-scoped, verifiable tasks such as fixed-format cleanup, common lookups, templated output, and preference capture. For high-risk actions, heavily context-dependent work, or one-off complex decisions, it should not default to bypassing checks. Keep habit-store.json auditable and retain a fallback path back to the full verification flow when rules misbehave.