Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/modelengine-group/fit-framework/refine-taskgit clone --depth 1 https://github.com/ModelEngine-Group/fit-frameworkWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00013 | $0.01930 |
| Opus 5 | $0.00006 | $0.00965 |
| Sonnet 5 | $0.00003 | $0.00386 |
| Haiku 4.5 | $0.00001 | $0.00193 |
Grade A, and why
refine-task scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 256 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Refinement Task Command
功能说明
处理代码审查中发现的问题,修复代码后重新进入审查流程。此命令用于 code-review 步骤发现需要修改的情况。
⚠️ CRITICAL: 状态更新要求
执行此命令后,你必须立即更新任务状态。参见规则 7。
执行流程
1. 验证前置条件
检查必需文件:
.ai-workspace/active/{task-id}/task.md- 任务文件.ai-workspace/active/{task-id}/review.md- 审查报告(或 review-supplement.md).ai-workspace/active/{task-id}/implementation.md- 实现报告
注意:{task-id} 格式为 TASK-{yyyyMMdd-HHmmss},例如 TASK-20260205-202013
如果任一文件不存在,提示用户先完成前置步骤。
2. 读取审查报告
仔细阅读审查报告(review.md 或 review-supplement.md 或 review-final.md),提取需要修复的问题:
问题分类:
- 🔴 必须修复(Blocker) - 阻塞问题,必须修复才能合并
- 🟡 建议修改(Major) - 重要建议,强烈推荐修复
- 🟢 优化建议(Minor) - 可选优化,可以考虑修复
提取信息:
- 问题标题
- 文件路径和行号
- 问题描述
- 修复建议
3. 使用 TodoWrite 规划修复任务
根据审查报告创建修复任务清单:
使用 TodoWrite 工具创建 todos:
- [ ] 修复问题 1: {问题标题}
- [ ] 修复问题 2: {问题标题}
- [ ] 修复问题 3: {问题标题}
...
优先级:
- 先修复所有 🔴 必须修复的问题
- 再修复 🟡 建议修改的问题
- 最后考虑 🟢 优化建议
4. 执行代码修复
按优先级逐个修复问题:
修复流程:
- 读取相关文件,理解问题上下文
- 按照审查建议修复代码
- 确保修复不引入新问题
- 在 TodoWrite 中标记该问题为已完成
修复原则:
- 严格按照审查建议修复
- 如果建议不明确,询问用户
- 如果发现新问题,一并修复
- 保持代码风格一致
5. 运行测试(如果有测试失败)
如果审查报告中提到测试问题:
# 运行单元测试
mvn test
# 运行特定测试
mvn test -Dtest=TestClassName
# 运行集成测试
mvn verify
确保所有测试通过后再继续。
6. 更新任务状态 (CRITICAL)
必须更新 .ai-workspace/active/{task-id}/task.md:
current_step: refinement
assigned_to: {当前AI,例如 claude}
updated_at: {当前时间,格式: yyyy-MM-dd HH:mm:ss}
在工作流进度中标记:
## 工作流进度
- [x] requirement-analysis (已完成)
- [x] technical-design (已完成)
- [x] implementation (已完成)
- [x] code-review (已完成 - 发现问题)
- [x] refinement (正在修复) ← 标记为进行中
- [ ] finalize (待执行)
7. 创建修复报告
创建 .ai-workspace/active/{task-id}/refinement-report.md,记录修复情况:
# 代码修复报告
## 修复概要
- **修复者**: {修复者}
- **修复时间**: {时间}
- **修复范围**: {修复的问题数量}
- **修复来源**: 代码审查反馈
## 修复内容
### 🔴 已修复的阻塞问题
#### 1. {问题标题}
**原问题**: {问题描述}
**修复方式**: {详细说明修复了什么}
**修改文件**: `{file-path}:{line-number}`
### 🟡 已修复的建议问题
#### 1. {问题标题}
**原问题**: {问题描述}
**修复方式**: {详细说明修复了什么}
**修改文件**: `{file-path}:{line-number}`
### 🟢 已采纳的优化建议
#### 1. {优化标题}
**原建议**: {建议描述}
**实施方式**: {详细说明如何实施}
**修改文件**: `{file-path}:{line-number}`
## 未修复的问题(如果有)
### {问题标题}
**原因**: {为什么没有修复}
**计划**: {如何处理}
## 测试结果
- [ ] 单元测试通过
- [ ] 集成测试通过
- [ ] 回归测试通过
- [ ] 新增测试(如果需要)
## 下一步
代码已修复,准备重新进入审查流程。
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 256 lines · 13 tokens per session scan A 142c1705f774
refine-task is a command published in the GitHub repository ModelEngine-Group/fit-framework (2,117 stars, last pushed 5mo ago), licensed MIT. It adds 13 tokens to every session and 1,930 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other commands, from other repositories
me-han-hu-draft
Draft a technical manuscript section in Han Hu manuscript mode using the private calibrated style corpus and evidence-preserving engineering workflow.
me-cfd-review
Review thermal-fluid CFD setup, boundary conditions, mesh, wall treatment, convergence, validation, and whether the claims are supported.
me-code-sanity
Run a fast preflight on thermal-fluid research code for units, baselines, leakage, physics checks, and result traceability.
me-correlation-check
Check thermal-fluid equations, empirical correlations, and dimensionless groups for validity range, assumptions, units, and claim strength.
me-experiment-plan
Plan thermal-fluid experiments with instrumentation, calibration, uncertainty, repeatability, heat-loss correction, operating envelope, and safety checks.
me-lit-matrix
Build a thermal-fluid literature matrix organized by mechanism, method, metric, validity range, benchmark value, and unresolved gap.