CRITICAL FIX: Changed 'if score >= 1.0' to 'if score > MAX_TASK_SCORE'
- Before: scores like 0.992 would pass through un-clamped
- After: scores exceeding 0.9 are properly clamped to 0.9
Also changed MAX_TASK_SCORE from 0.99 to 0.9 for safety margin.
fix: ensure all task scores are strictly between 0 and 1 (0.1-0.99 range)
Enforce strict in-range scoring via shared clamp, align task max_steps with openenv metadata, and update docs/tests accordingly.
Co-Authored-By: Oz <oz-agent@warp.dev>
fix: adjust score range to (0, 1) for Phase 2 validation
