Legacy DSH skill report
DSH plugin validation
controls only. Large plugin audit cases are useful controls; no stable correctness separation from the without arm.
8/8 bothplugin audit
7/7 bothsupplychain
7/7 bothroute-authz
8/8 bothmultifinding
Disabled: Legacy cases are controls only; hidden from default skill discovery until a discriminating case is frozen. The skill is kept in the repository as
SKILL.md.disabled.No replicated correctness signal. Treat as controls until a new discriminating case is frozen.
Documented comparisons
legacy evidence, not sealed records| Case / batch | With skill | Without skill | Reading |
|---|---|---|---|
| large-plugin-audit (204 files) | 8/8 | 8/8 | large control, not discriminator |
| large-plugin-supplychain (205 files) | 7/7 | 7/7 | efficiency hint only |
| large-route-authz at 80 | 7/7 | 6/7 then 7/7 | training case candidate; holdout only budget difference |
| large-plugin-multifinding at 150 | 8/8 | 8/8 | large efficiency control |
Source and limitations
These numbers come from benchmarks/evaluation/TASK_LEVEL.md and benchmarks/evaluation/FINAL_REPORT.md. Raw trial records were not retained as sealed records in this rebuild; counts here are documented legacy results.
No cross-skill or cross-model aggregate is computed. The attribution report is the only page in this site backed by the current sealed evidence set.