T-mn816iew.json 1.53 KB

Edit Raw Blame History



1

2

3

4

5

6

7

8

9

10

11

12

{
  "taskId": "T-mn816iew",
  "from": "innovator",
  "to": "evolver",
  "task": "【FDE Phase 3 - data-analyst 技能进化训练】<br><br>## 输入资料<br>- 知识体系: data/workspace/data-analyst/digest/knowhow.md<br>- 评分标准: data/workspace/data-analyst/digest/benchmark.md (pass_score = 90)<br>- 训练集: data/workspace/data-analyst/digest/train.json<br>- 测试集: data/workspace/data-analyst/digest/test.json<br>- 初始 SKILL: data/workspace/data-analyst/digest/SKILL.md<br><br>## 训练要求<br>1. 以 SKILL.md 为起点，继续迭代优化<br>2. 硬性及格线: 总分 >= 90 (100分制)，低于此分数不能标记为完成<br>3. 最大迭代次数: 10 轮<br>4. 每轮基于测试集评分，分析弱维度，针对性修改 Skill 定义<br><br>## 产出要求<br>- 产出目录: data/workspace/data-analyst/evolve/<br>- 文件1: skill_data_analyst.md (迭代后的技能定义)<br>- 文件2: evolution_log.json (训练记录，每轮评分和修改说明)<br><br>## 部署要求（训练达标后必须执行）<br>1. 将最终 SKILL.md 复制到 data/companies/abc/agents/data-analyst/.claude/skills/data-analysis/SKILL.md<br>2. 更新 data/companies/abc/agents/data-analyst/.claude/skills/skills.md 添加技能清单<br>3. 更新 data/companies/abc/agents/data-analyst/config.json 中 status 为 ready<br><br>完成训练后，汇报最终评分和部署状态。",
  "priority": "P1",
  "goalId": null,
  "nodeId": null,
  "executionConvId": "task-T-mn816iew",
  "createdAt": "2026-03-26T22:15:17.916Z",
  "status": "dispatched"
}