Eval directory

Evals for StaffScheduleCare

1 evaluation pack covering adversarial robustness, safety gates, workflow quality, and operator-level checks for StaffScheduleCare AI products.

Document Agents

Use evals for StaffScheduleCare

Available eval packs for StaffScheduleCare

1 pack ready to run.

Ssc Parser Regression V1
Regression eval library for SSC's deterministic XLSX intake-form parser. Each fixture is a small synthetic workbook designed to lock in one specific parser code path tied to a known/resolved bug — when the parser regresses on that path, the fixture name tells you which fix broke. This dataset contains the 11 synthetic regression fixtures only. SSC's local CI runs an extended dataset that also includes their 5 real client fixtures (kept in the SSC repo for data-isolation reasons). EXECUTION MODEL — EXTERNAL ONLY (until Phase 2): This spec is designed to run via the SSC SDK in the SSC repo's CI, which invokes the TypeScript parser locally and reports per-example results to Corsac. The dataset shape ({input: {fixture_path}, expected: {data, errors, warnings}}) reflects that workflow — it does NOT match the doc_agent runner's input/output contract (input.document/input.instruction → {result: ...}). Triggering "Run" from the Corsac UI will currently fail with structural mismatches; that's expected. Phase 2 work fixes this two ways: (a) adds a 'parser_regression' category whose runner skips server-side execution, and (b) ships the external-run endpoint that accepts pushed results from SSC's CI. Tracked in issues/phase-2-ssc-pilot-readiness.md (2-09, 2-17) and issues/phase-1.5b-autogen-hardening.md.

Why eval StaffScheduleCare AI

StaffScheduleCare's AI features ship behind brand promises about accuracy, safety, and reliability. Buyers and integrators need to know those promises hold up under adversarial prompts, edge-case workflows, and the long tail of real customer inputs — not just the demo path.

The Corsac eval library for StaffScheduleCare measures four dimensions teams care about most when deploying document agents agents:

Adversarial robustness — does the agent resist prompt injection, jailbreaks, and social-engineering attempts?
Workflow quality— does it complete the task buyers were shown in the demo, on inputs that don't look like the demo?
Safety gates — does it escalate or refuse when it should, and only then?
Operator quality — does it preserve analyst trust by surfacing the right context at the right time?

Every eval pack above is hand-authored against StaffScheduleCare's public product surface and runnable in Corsac with your own data.