83 lines
2.8 KiB
Markdown
83 lines
2.8 KiB
Markdown
# 102 Full Sweep Improvement Coverage Delta Report
|
|
|
|
> Date: 2026-04-19
|
|
> Roadmap: `docs/superpowers/plans/2026-04-19-102-full-sweep-improvement-roadmap-plan.md`
|
|
> Baseline: `tests/fixtures/generated_scene/full_sweep_dry_run_2026-04-19.json`
|
|
> Follow-Up: `tests/fixtures/generated_scene/full_sweep_improvement_followup_2026-04-19.json`
|
|
|
|
## Headline Delta
|
|
|
|
| Metric | Baseline | Follow-up | Delta |
|
|
| --- | ---: | ---: | ---: |
|
|
| Dry-run auto-pass | 40 | 48 | +8 |
|
|
| Dry-run actionable coverage | 66 | 96 | +30 |
|
|
| Timeout / source-unreadable | 31 | 2 | -29 |
|
|
| Misclassified | 5 | 4 | -1 |
|
|
| Fail-closed-known | 26 | 48 | +22 |
|
|
|
|
The increase in `fail-closed-known` is expected. The roadmap converted many timeout/no-report cases into structured fail-closed outputs instead of process-level failures.
|
|
|
|
## Follow-Up Status Counts
|
|
|
|
| Status | Count |
|
|
| --- | ---: |
|
|
| `auto-pass` | 48 |
|
|
| `fail-closed-known` | 48 |
|
|
| `misclassified` | 4 |
|
|
| `source-unreadable` | 2 |
|
|
| `missing-source` | 0 |
|
|
| `unsupported-family` | 0 |
|
|
| Total | 102 |
|
|
|
|
## Remaining Timeout Cases
|
|
|
|
Only `2` timeout cases remain:
|
|
|
|
1. `任务报表`
|
|
2. `力禾动环系统巡视记录`
|
|
|
|
These are not known-family samples.
|
|
|
|
## Remaining Misclassification Cases
|
|
|
|
`4` route conflicts remain:
|
|
|
|
| Scene | Expected group | Inferred archetype |
|
|
| --- | --- | --- |
|
|
| `95598报修工单日管控` | `G3` | `host_bridge_workflow` |
|
|
| `95598重要服务事项报备统计表` | `G3` | `host_bridge_workflow` |
|
|
| `台区线损台区月度高负损预测` | `G2` | `host_bridge_workflow` |
|
|
| `配网支撑月报(95598抢修统计报表)` | `G3` | `host_bridge_workflow` |
|
|
|
|
One previous `G1-E` conflict, `高低压新增报装容量月度统计表`, now returns to `single_request_enrichment` and auto-passes.
|
|
|
|
`用电报装信息统计列表` now returns to `single_request_enrichment` but remains fail-closed, so it is no longer a route-overpreference case.
|
|
|
|
## Structured Fail-Closed Result
|
|
|
|
The baseline had `25` no-report failures. The follow-up emits structured fail-closed reports for these paths, with:
|
|
|
|
1. inferred archetype
|
|
2. `generationStatus = fail-closed`
|
|
3. `failureStage = readiness-before-report`
|
|
4. blocker reason
|
|
5. readiness missing pieces and risks
|
|
|
|
This makes failures machine-readable without weakening gates.
|
|
|
|
## Interpretation
|
|
|
|
The roadmap delivered measurable improvement:
|
|
|
|
1. timeout noise was reduced from `31` to `2`
|
|
2. auto-pass increased from `40` to `48`
|
|
3. actionable coverage increased from `66` to `96`
|
|
4. no-report failures were converted into structured fail-closed results
|
|
|
|
Remaining work is no longer the same broad ambiguity. The next blocker is the smaller route-conflict set, especially `G3/G2` vs `host_bridge_workflow`.
|
|
|
|
## Stop Statement
|
|
|
|
This report does not update `scene_execution_board_2026-04-18.json` and does not promote any scene.
|
|
|