SIGN IN SIGN UP

feat(scenarios): add Phase 1 validation scenarios for 3-wave matrix

Adds 9 scenario JSON files matching the Phase 1 model matrix:

  Wave 2 (Sonnet 4.6):
    - prompt-cache-reuse         cache_control placement, sticky-on betas
    - mcp-tool-call              MCP tool naming, defer_loading, alphabetization
    - tool-search-defer          defer_loading + tool_search_tool server tool
    - long-context               clear_tool_uses_20250919 context management
    - long-context-compact       compact_20260112 (Sonnet/Opus exclusive)
    - adaptive-thinking-shared   adaptive thinking + effort (shared with Opus)

  Wave 3 (Opus 4.7 exclusive):
    - opus-thinking-display-override   display: 'omitted' default observation
    - opus-effort-xhigh                effort: 'xhigh' observation
    - opus-rejects-manual-thinking     type: 'enabled' always-adaptive regression

Wave 1 scenarios (minimal-hi, append-system-prompt, oauth-token-refresh)
already exist in-tree. All 12 scenarios use the VerificationScenario schema
(id, name, description, prompt, expectedBehavior, requiredFields, runnerMode,
requestPathContains). Evidence-only; no runtime behavior change.

Unblocks Phase 1 Task 1.2 per
.sisyphus/plans/griffin-parity-and-cc-optimizations.md Task 1.1.5.
V
Vacbo committed
4fe1aabcd8aab037d6746365af35891e45f9b4ec
Parent: 49dfd0b