← 返回
未分类 中文

Task Retrospective

Conduct structured self-evaluations after tasks to analyze efficiency, accuracy, approach quality, and extract patterns for improved future performance.
在任务结束后进行结构化自评,分析效率、准确性、方法质量并提炼模式,以提升未来表现。
charlie-morrison
未分类 clawhub v1.0.1 1 版本 100000 Key: 无需
★ 0
Stars
📥 381
下载
💾 1
安装
1
版本
#latest

概述

Task Retrospective

Structured self-evaluation for AI agents after completing tasks. Analyze what worked, what failed, and extract reusable patterns to improve future performance. Use after completing complex tasks, debugging sessions, or multi-step workflows.

Usage

Run a retrospective on the task I just completed.

Or with specific context:

Retrospective: [task description]. 
Outcome: [success/partial/failure].
Time spent: [duration].
What surprised me: [unexpected findings].

How It Works

  1. Reconstruct — review the task timeline (steps taken, tools used, decisions made)
  2. Evaluate — score each phase on efficiency, accuracy, and approach quality
  3. Extract — identify reusable patterns, anti-patterns, and decision heuristics
  4. Record — generate a structured retrospective for future reference

Evaluation Dimensions

Efficiency

  • Were there unnecessary steps or dead ends?
  • Could tool calls have been batched or parallelized?
  • Was the research phase too long or too short?

Accuracy

  • Was the final output correct and complete?
  • Were there false starts or incorrect assumptions?
  • Did the solution match the actual requirements?

Approach Quality

  • Was the problem decomposition effective?
  • Were the right tools chosen for each step?
  • Would a different strategy have been faster?

Learning Extracted

  • What new patterns can be reused?
  • What anti-patterns should be avoided?
  • What domain knowledge was gained?

Output Format

## Task Retrospective

### Summary
[1-2 sentences: what was the task, what was the outcome]

### Timeline
| Phase | Duration | Verdict |
|-------|----------|---------|
| Research | Xm | Efficient / Too long / Insufficient |
| Planning | Xm | Good / Skipped / Over-planned |
| Execution | Xm | Clean / Had rework / Multiple attempts |
| Validation | Xm | Thorough / Skipped / Caught issues |

### What Worked
- [Pattern that should be repeated]

### What Didn't Work
- [Anti-pattern to avoid] → [Better alternative]

### Reusable Patterns
- **Pattern name**: [Description of when and how to apply]

### Key Decisions
- [Decision point] → [Choice made] → [Outcome: good/bad/neutral]

### Improvement Actions
- [ ] [Specific action to improve future performance]

Advanced Usage

Compare Approaches

Compare my approach to [task] with the ideal approach. 
What I did: [steps].
What I should have done: [if known].

Pattern Library

Over time, retrospectives build a pattern library:

Review my last 5 retrospectives. What recurring patterns emerge?
Which improvement actions have I actually followed through on?

Team Retrospective

Run a retrospective on this multi-agent workflow.
Agents involved: [list].
Handoff points: [where work transferred between agents].
Bottlenecks: [where things slowed down].

版本历史

共 1 个版本

  • v1.0.1 当前
    2026-05-07 12:53 安全 安全

安全检测

腾讯云安全 (Keen)

安全,无风险
查看报告

腾讯云安全 (Sanbu)

安全,无风险
查看报告

🔗 相关推荐

ai-agent

Find Skills

root
帮助用户发现和安装智能体技能,当用户询问如「如何做X」、「找X的技能」、「有能做...的吗」等问题时
★ 1,521 📥 578,203
ai-agent

Agent Browser

rez0
用于 AI 代理的浏览器自动化 CLI。当用户需要与网站交互(包括浏览页面、填写表单、点击按钮、截图等)时使用。
★ 868 📥 347,443
dev-programming

Devcontainer Validator

charlie-morrison
在 VS Code 开发容器中验证 devcontainer.json 的语法、结构、功能、端口、生命周期脚本、定制项及安全最佳实践。
★ 0 📥 560