Use at the start of every ClawForce development session to drive the self-improvement loop. Triggers on "clawforce dev", "dev loop", "improve clawforce".
Drive ClawForce development by inspecting health, identifying issues, implementing fixes, and verifying via the live system.
Run these commands:
cd ~/workplace/clawforce && npx vitest run 2>&1 | tail -10
cd ~/workplace/clawforce && npx tsc --noEmit 2>&1 | tail -10
grep -i "clawforce.*error\|clawforce.*warn\|clawforce.*fail" /tmp/openclaw/openclaw-*.log 2>/dev/null | tail -10grep -rn "TODO\|FIXME" ~/workplace/clawforce/src/ | wc -l
ls ~/workplace/clawforce/docs/superpowers/specs/
Pick the single highest-priority item.
Read the relevant source. Write a 2-3 sentence plan. Note downstream impacts.
Dispatch cf-worker with a targeted task:
openclaw agent --agent cf-worker --message "[task that exercises the fix]"
Check results in the DB.
Success = tests pass, types clean, verification dispatch compliant. If yes → loop to Step 1. If no → debug and re-verify.
PyTorch深度学习模式与最佳实践，用于构建稳健、高效且可复现的训练流程、模型架构和数据加载。