Prep my visit skill

[pswider]

Summary:

Add a new prep-my-visit skill that generates an IPS-aligned pre-visit package (provider summary, patient companion, lab opportunities, and portal-ready snippets) with strict health-data-only boundaries.
Add deterministic validation helpers for IPS structure, lab-opportunity guardrails, and snippet constraints under skills/prep-my-visit/scripts/.
Add a comprehensive Waza eval suite for prep-my-visit (positive, negative/routing, guardrail, adversarial, and deterministic golden scenarios), plus a dedicated workflow for skill readiness checks.
Add a detailed skill README with practical usage guidance for Claude/OpenClaw runtime and Copilot-backed Waza eval execution.

What’s included
New skill
skills/prep-my-visit/SKILL.md
skills/prep-my-visit/README.md
skills/prep-my-visit/references/*
skills/prep-my-visit/scripts/*
New eval suite
evals/prep-my-visit/eval.yaml
evals/prep-my-visit/eval.mock.yaml
evals/prep-my-visit/README.md
evals/prep-my-visit/tasks/*.yaml
evals/prep-my-visit/fixtures/golden-visit-input.json
CI/workflow
.github/workflows/prep-my-visit-evals.yml

Safety and scope guarantees
No diagnosis, prognosis, or treatment recommendations.
No auto-ordering tests and no auto-send portal behavior.
No billing/insurance/prior-auth/EOB domain content.
PHI boundary preserved to workspace-only operations.
Category B lab suggestions constrained to discuss-with-doctor posture.
Category C (DTC path) requires explicit opt-in.

Test plan

waza check skills/prep-my-visit

waza run evals/prep-my-visit/eval.mock.yaml -v (structural/demo lane)

waza run evals/prep-my-visit/eval.yaml -v (live lane; pending Copilot quota availability)
Notes
This PR is intentionally scoped to prep-my-visit only (skill + evals + workflow + docs), separated from larger branch work for clean review and safer merge.

[pswider]

eady to merge from my side.

Final pre-merge status:

Scope is clean and intentional: prep-my-visit skill + eval suite + dedicated workflow + docs only.
Safety boundaries are explicit and enforced in both skill instructions and eval tasks (no diagnosis/treatment, no auto-send, no billing/insurance drift, PHI workspace boundary).
Deterministic validation helpers are included for IPS structure, lab-opportunity guardrails, and snippet constraints.
Mock/structural eval lane is in place for no-quota validation.
Known release note:

Live-model Waza run is still pending Copilot usage availability:
waza run evals/prep-my-visit/eval.yaml -v
Given current constraints, this is merge-ready for controlled rollout with the live-eval follow-up tracked immediately when quota is available.