WorkBuddy Bench（腾讯官方评测集）

WorkBuddy Bench is a benchmark for evaluating coding agents on real-world work, reverse-engineered from actual developer, PM, algo, QA, ops, and security tasks. Given a task and a sandboxed workspace, an agent is asked to produce the correct change (a patch, an artifact, a report) and is graded against a test suite.

课时

• .agents
  • skills
    • wbbench-report-skills
      • references
        • WB-Bench-Code Report Workflow
        • WorkBuddyBench-Office Report Workflow
        • WB-Bench-Web Report Workflow
      • WB-Bench Report Skills
    • wbbench-run-setup
      • references
        • Phase 0 — Datasets
        • Phase 1 — Environment
        • Phase 2 — Model
        • Phase 3 — Credentials (.env)
        • Phase 4 — Job
        • Phase 5 — Launch
        • Phase 6 — Metrics + trajectory analysis
      • WB-Bench Run Setup
• configs
  • harnesses
    • claude-code
      • claude-code harness
    • codebuddy-code
      • CodeBuddy Code (cbc) harness
    • Adding / adjusting a harness
  • Configuration system
  • 配置体系
• datasets
  • Datasets
• README
• README.zh

开始学习 → WB-Bench-Code Report Workflow