Skip to content

Memory Effectiveness and ROI Scorer with Auto-Pruning #405

@github-actions

Description

@github-actions

Strategic Improvement

The memory system stores failure patterns and learnings, but we don't measure whether they actually help. Build a scorer that correlates memory entries with pipeline outcomes: did this pattern prevent a failure? reduce iterations? improve success rate? Auto-prune ineffective patterns and surface high-value learnings. Complements #373 (memory decomposition) with a feedback loop that validates the learning system works.

Acceptance Criteria

  • Score each memory entry: correlation with success rate, iteration reduction, failure prevention
  • Track memory entry usage: how often injected, how often agent acted on it, outcome delta
  • Add effectiveness metrics to shipwright memory show: score (0-100), usage count, impact
  • Auto-prune entries with score <20 after 10+ pipeline runs (low-value noise)
  • Generate monthly report: "Top 10 most effective patterns" and "Ineffective patterns removed"
  • Emit memory effectiveness events to enable cross-pipeline learning optimization

Context

  • Priority: P2
  • Complexity: standard
  • Generated by: Strategic Intelligence Agent
  • Strategy alignment: P2: Intelligence & Learning

Metadata

Metadata

Assignees

No one assigned

    Labels

    auto-patrolCreated by autonomous patrol agentsready-to-buildIssue is ready for autonomous pipeline processingstrategicCreated by strategic intelligence agent

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions