WS4 Part D: golden-set cases for the five new UK templates + Disclosure Review
Adds 22 new eval case files (17 citation, 5 judged) and 5 new rubric files, covering all five new WS4 templates plus the renamed Disclosure Review (which gets its own judged rubric per the three-way privilege split); the Companies House Due-Diligence Snapshot template reuses the existing jud-due-diligence-snapshot.yaml + rubric per the brief and is extended with 2 new citation cases rather than duplicated. Every citation case's input.text is a short realistic excerpt of what the matching template instructs the model to produce, using the exact statutory forms the templates mandate (s.N Act Year / SI year/number). None are marked smoke (the smoke subset is already at its 5-case cap). Judged cases are marked pending (no workflow-fixture output yet) with real rubrics ready for when fixtures land. Verified: `npm run evals` from the worktree root - 20 passed, 0 failed, 1 skipped (Companies House deterministic case, no API key set - expected), 6 pending (all judged cases). Every new citation case resolved live against legislation.gov.uk with zero failures. `npm run evals:smoke` unchanged (3 passed, 1 skipped). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
| Repository | Sloth-ninja/JessicaOSS |
|---|---|
| Author | sloth-ninja <ezanahaddis@gmail.com> |
| Authored | |
| Committed | |
| Parents | 0e595215 |
| Stats | 27 files changed , +434 |
| Part of | WS4: UK terminology sweep + five new UK workflow templates |
Capture this commit into my fork
Download a Markdown prompt that tells Claude how to port this
exact commit into your working tree. Run it via
claude -p < capture-commit-de65ad29.md
from inside the repo you want the change in.