Letthedreamersrise/wallet-rescue-ops

Increase task max_steps for validator agent headroom

Switch to mandatory key=value log format, add /grader endpoint, strip action amounts from output

Fix double-normalization: use ScoreBreakdown.normalized() classmethod, remove from observation, fix inference.py

Normalize environment rewards to (0.001, 0.999) - Phase 2 validator requires all rewards in (0,1)

Normalize all step rewards to (0,1) range, remove reward_total from END logs

Clamp scores to strict (0.001, 0.999) range - validator rejects 0.0 and 1.0

Define explicit main() in server/app.py for openenv validate

Add uv.lock and server/app.py for openenv validate compatibility

Match exact env var pattern from submission checklist: defaults for API_BASE_URL/MODEL_NAME only, no default for HF_TOKEN

fix: open external links in new tab to avoid iframe CSP errors

Add landing page at root URL for judges visiting the Space

fix: use valid HF Space colorTo value

Wallet Rescue Ops v2: OpenEnv incident-response environment with 5 graded tasks, 12 scenarios, 6-component reward, heuristic baseline, 19 tests, HF Spaces Docker ready