Test Coverage: agent-loops provider diagnostics

[NickCrew]

Module information

Files: skills/agent-loops/scripts/specialist-review.sh, skills/agent-loops/scripts/diff-test-audit.sh
Tests: tests/unit/test_agent_loops_review_scripts.py

Gap

The agent-loops effort-control work covers command construction, default models, optional effort handling, secondary effort precedence, and Codex diagnostics. The final diff audit left two P2 gaps in diagnostic-message coverage:

• specialist-review.sh needs explicit Claude and Gemini cases that verify the selected provider's model or effort message and reject messages from other providers.
• diff-test-audit.sh needs the same Claude and Gemini diagnostic matrix.

Execution behavior is already covered. This follow-up is limited to operator-facing stderr output.

Acceptance criteria

• Explicit Claude runs verify the configured effort message in both entrypoints.
• Explicit Gemini runs verify the configured model message in both entrypoints.
• Each case verifies that diagnostics from inactive providers are absent.
• The focused provider-script test suite passes.