design-loops

Designing compounding feedback loops requires mapping action-to-data-to-analysis-to-better-action chains, specifying degradation rules for varying data volumes, and defining explicit inter-loop connections to improve agent OS performance over time.

Handling cold-start and sparse data scenarios requires defining explicit degradation rules and thresholds within your feedback loop design, ensuring the agent OS prevents degraded performance when operational data volumes are insufficient.

Mapping data producers and readers across nested feedback loops involves specifying complete loop design specifications that trace coaching signals and actions, ensuring explicit inter-loop connections and metrics are defined in the retro model.

Defining explicit metrics in the retro model is required to track bottlenecks and measure cross-loop impact, ensuring loop outputs correctly influence other connected loops and accumulate operational data effectively.

Configuring feedback loop design files involves writing the complete loop specifications to domain-input/feedback-loops.md and output/designs/loop-designs.md, then running the /generate-os command to assemble the system architecture.

Agent OS performance degrades during data accumulation when the feedback loop design lacks explicit degradation rules for sparse or rich data scenarios, or when inter-loop connections and retro model metrics are not properly specified.