[rkboyce]

Purpose

Incremental upstream review for the conference-focused Study Agent integration work.

Included so far

• Review-gated concept-set and multi-component cohort-plan orchestration.
• Atlas3/WebAPI integration contract, provenance, and deployment documentation.
• Config wizard optional-section guidance and secret-name handling tests.
• Explicit LLM endpoint/model configuration errors.

Companion repositories

Atlas3 and WebAPI changes are maintained on matching conference branches and are documented in docs/ATLAS_WEBAPI_INTEGRATION.md. This PR remains service-focused.

Validation

• Focused ACP/client tests: 56 passed.
• Config and wizard tests: 31 passed.
• Ruff and whitespace checks passed for the changed Python.

Review plan

This remains draft while symposium stabilization continues. New commits will be small and coherent; risky architectural moves are tracked separately in issues #98, #99, and #100.

[rkboyce]

Final validation on head 80d42c2: CodeQL is successful. uv run --extra dev python -m pytest tests completed with 350 passed and 3 failures in test_r_phenotype_conversion_handoff_static.py. Those three assertions read the ignored, separately checked-out sandbox/slashOhdsiStrategusAssistant R tree rather than files tracked by this repository; its current text has drifted from the static expectations. No PR-tracked StudyAgent source failed. The all-path discovery run also attempted to collect the separately vendored ChatKit repository under sandbox/, which has its own isolated dependencies; the scoped tests/ run excludes it.\n\nThis PR is ready for merge under the active main ruleset.