kaizen

Show kaizen usage guide, available profiles, examples, and configuration tips

Show improvement run history and KPI trends for a kaizen profile

Run a kaizen improvement loop with a specified profile

Interpret kaizen KPI measurements by comparing current values against baseline and history. Identifies trends, calculates deltas, performs root-cause analysis, ranks improvement opportunities, and flags anomalies. Dispatched during ANALYZE phase of each kaizen iteration.

Execute kaizen measurement tools and collect KPI values. Runs auto-generated Python or TypeScript measurement scripts, captures JSON output, and writes structured results to the iteration directory. Dispatched during MEASURE and VERIFY phases of each kaizen iteration. Lightweight and fast — optimized for frequent invocation.

Generate concrete, minimal improvement proposals for kaizen iterations. Reads analysis and hypotheses, respects mutation boundaries, avoids repeating reverted approaches, and produces actionable change plans with expected KPI impact estimates. Dispatched during PROPOSE phase.

Adversarial validation agent for kaizen improvement loops. Reviews measurement tools for correctness and gaming vulnerability, validates that improvements are genuine and aligned with the profile mission, and checks immutability boundary compliance. Dispatched at BOOTSTRAP for tool review and as final gate after loop completion. Uses deep reasoning to catch subtle issues.

Recursive improvement loop engine inspired by karpathy/autoresearch. Orchestrates continuous improvement via Sequential Thinking MCP with 8-phase iterations (MEASURE, ANALYZE, HYPOTHESIZE, PROPOSE, APPLY, VERIFY, DECIDE, LOG). Supports greedy and multi-objective ratcheting strategies with configurable autonomy levels. Use when the user says 'run kaizen', 'kaizen loop', 'improve', 'optimization loop', 'continuous improvement', 'recursive improvement', 'iterative optimization', 'run improvement profile', or wants to iteratively improve code, configuration, or processes against measurable KPIs. Also activates when user references a specific profile name like 'claude-code-usage', 'code-refactoring', or 'process-improvement'. Requires **~~sequential-thinking** connector.

Interactive workflow to create custom kaizen improvement profiles. Guides users through KPI definition, data source identification, mutation scope selection, autonomy configuration, and profile validation. Use when user says 'create kaizen profile', 'design improvement loop', 'new kaizen profile', 'custom profile', 'custom kaizen', 'define improvement target', or wants to create a new improvement loop for a domain not covered by the bundled profiles.

View improvement history and KPI trends from kaizen runs. Reads .kaizen/runs/ audit trail to show cross-run improvement trajectories, diminishing returns detection, and formatted KPI reports with delta tables. Use when user says 'kaizen report', 'kaizen history', 'improvement history', 'show kaizen results', 'KPI trends', 'how is improvement going', 'kaizen status', or wants to review past improvement loop outcomes.