marsai:dev-feedback-loop

Automating development cycle feedback involves calculating assertiveness scores, evaluating agent prompt quality, and aggregating cycle metrics to generate improvement reports. This process captures gate transitions, failures, and successes to provide actionable insights for future software engineering work.

Cycle metrics collection works by automatically evaluating prompt quality across all participating agents and computing assertiveness scores. It aggregates data from gate transitions, validation outcomes, and task completions to assess overall development cycle health.

The best way to track gate transitions and failures is to apply automated metrics collection after task completion and validation outcomes. This captures successes and failures at the end of a development cycle, enabling comprehensive root-cause analysis and targeted improvements.

To generate improvement reports, apply feedback automation after a development cycle ends to calculate assertiveness scores and evaluate prompt quality. The consolidated reports are then saved under docs/feedbacks/cycle-{date}/ for archival and action planning.

You should run a development cycle quality assessment after task completion, after validation outcomes, and at the end of a development cycle. This timing ensures accurate capture of gate transitions, failures, and successes to drive continuous improvement.