Add dynamic text model discovery to public Skill

[erickkkyt]

Summary

• add a four-step setup fast path without hard-coded model names
• define the complete model inventory as the live union of /v1/models and /v1/media/models
• add an authenticated text-model List -> choose -> call flow using /v1/responses
• clarify that capability Search currently covers image/video generation models, not text models
• extend the read-only setup walkthrough and tests to verify both model catalogs

Verification

• npm run verify
• system skill validator
• git diff --check
• scoped secret-pattern scan

Not included

• no production publish or deployment
• no gateway-side change to merge text models into /v1/capabilities/search; the Skill uses /v1/models as the authoritative text-model source