工作队列与优先级排序
Work Queue & Prioritization
智能工作队列系统，自动排序和聚类代码问题，控制信息密度
子问题
1.Multi-dimensional finding ranking
2.Auto-clustering related findings
3.Plan operations (focus/skip/defer/move)
4.Noise budget for information density control
5.Plan-State reconciliation after scan churn
6.Cluster collapse for reduced cognitive load
7.Subjective vs mechanical finding segregation
8.Stale skip auto-resurfacing by scan count
各项目的解法1 solutions
Signals
横向对比
| 维度 | Desloppify |
|---|---|
| 排序策略 | 四层 Tier + confidence + review_weight + count 多维 tuple 排序 |
| 噪声控制 | per-detector cap + global round-robin 双层预算 |
| 聚类机制 | detector/subtype/file 三维分组键自动聚类 + cluster 折叠 |
| 计划操作 | move/skip/focus/describe/annotate/reset 六种 plan 操作 |
| 状态协调 | reconcile 自动检测消失 findings + 90 天 superseded TTL 清理 |
| 主观降级 | subjective findings 强制 T4 + 二级排序位降级，不与机械 findings 竞争 |
最佳实践
1.Round-robin global budget for fair detector representation
2.Tier-weighted scoring for impact-based prioritization
3.Tuple-based multi-key sort for explainable ranking
4.User-modified flag to protect manual cluster edits from auto-regeneration
5.Superseded record with TTL for graceful finding lifecycle