[EPIC] Automated quality gates

[ypldan]

Goal

Make model-support and system-correctness guarantees enforceable by CI instead of by tribal knowledge. Integration tests exist but run only locally (#270); new models are added to them ad-hoc with no documented process or support guarantee (#271); the core+quickapps system path is barely covered because current tests exercise the orchestration/LLM layer (#317); and 584 pyright errors sit outside the lint gate (#398). The common goal: a contributor (or an agent) can trust green CI to mean "supported models still work and the system still integrates."

Scope — child issues

• Integration tests update process + supported models guarantees #271
• Integration tests to CI #270
• Integration tests for internal system(core+quickapps) #317
• Fix pyright errors and warnings #398

Out of scope

• E2E test expansion beyond the existing make e2e_test scope.
• Test-authoring developer tooling (PR docs(skills): add quickapps-write-integration-test skill #350's skill) — supportive, not gating.

Dependencies and sequencing

• #271 first — the process and guarantees define what CI must enforce; #270 then makes it enforceable (needs a deployment/credentials strategy for CI); #317 extends coverage to the system path using the same harness.
• #398 is independent — schedule as capacity allows; it only shares the "green means trustworthy" goal.
• PR chore: auto-sharing prompt and contexts test case #341 (test-harness support for contexts/skills) and PR docs(skills): add quickapps-write-integration-test skill #350 (test-writing skill) support this epic and should land under it.

Additional information

Related: PR #341, PR #350, .claude/rules/integration-testing.md (harness mechanics).