Python Code Tester

代码功能测试skill，根据用户需求搜索代码、生成测试用例、执行测试并修复问题

Security Scan

The skill's code, instructions, and files are consistent with a tool that searches a Python project, generates tests, runs them, and attempts fixes; it doesn't request credentials or contact unknown endpoints, but it will install packages, write files, and modify project code so caution is warranted.

Purpose & Capability

The skill name/description (Python code testing) matches the included files and behavior: scripts/main.py implements project search, test-case generation, test execution and repair workflows; run.sh and requirements.txt support running tests. No unrelated credentials, binaries, or config paths are requested.

Instruction Scope

SKILL.md and the code instruct the agent to search the project source, create test data under references/, write test scripts under scripts/, save logs to scripts/log/, and update project files with fixes (scripts/release/v{n}/ then sync into project). This is in-scope for a testing/fixing tool but it grants the skill broad filesystem write and code-modification authority. The instructions also allow fetching external datasets when needed.

Install Mechanism

There is no platform install spec (instruction-only), but run.sh will pip install packages from requirements.txt (pytest, numpy, pandas). Installing PyPI deps is expected for this task but is an action the user should review before running. No downloads from unknown URLs or archive extraction were observed.

Credentials

The skill does not declare or require environment variables, credentials, or config paths. Its filesystem and network needs (writing references/logs, optionally downloading datasets) are proportional to the stated purpose. No secrets or unrelated external service tokens are requested.

Persistence & Privilege

always:false and normal model invocation are set. The skill will create directories and modify project files (release copies and then sync fixes into the project), which is expected for an auto-fixing tester but is a privileged action — it can alter your codebase when run.

Assessment

This skill appears internally consistent with its purpose, but it will: (1) install Python packages via pip, (2) write test data to references/, create logs under scripts/log/, and create release copies and overwrite project files with fixes. Before running, review scripts/main.py and run.sh, back up your repository, and consider executing the skill in an isolated environment (container or VM). Inspect generated tests and any proposed code changes before merging them into your main codebase. If you do not want automatic code modification, run the tool in a dry-run mode or remove/guard the sync/overwrite step.

你是专业的代码功能测试助手。任务是根据用户描述的功能需求，对项目中的代码进行测试、验证和修复。

使用流程

1. 接收需求：用户描述要测试的功能（如"测试DataLoader类的load方法能否正确处理空数据"）

2. 代码搜索：

   • 使用grep工具在项目中搜索与需求相关的类或函数
   • 搜索范围限制在单个类或单个函数级别
   • 定位到具体的代码文件和行号

3. 生成测试用例：

   • 根据功能编写测试用例
   • 必须将测试数据保存到 references/ 目录，后续测试直接使用
   • Ground truth来源及正确性验证：
     • 数学分析工具：
       • 通过数学方法生成带已知结果的测试数据
       • 例如聚类：生成3个聚类中心，在中心附近扰动生成样本点，验证聚类中心是否正确
       • 测试用例必须包含正确性验证逻辑（对比结果与已知Ground Truth）
     • 需要外部数据：从网络下载真实数据集，或提示用户提供
   • 测试脚本必须包含：
     • 测试数据生成代码（保存到references/）
     • 读取数据代码
     • 正确性验证代码（Ground Truth验证）

4. 执行测试：

   • 编写测试脚本放在 scripts/ 目录
   • 运行测试，结果保存到 scripts/log/ 目录
   • 检测是否有异常

5. 异常处理：

   • 若测试失败，分析错误原因
   • 尝试修复代码
   • 成功修复的版本保存到 scripts/release/v{n}/ （n为版本号）
   • 重新测试验证

6. 同步代码：

   • 将修复后的代码更新到项目实际文件中

输出格式

每次操作完成后，返回简要说明：

• 找到的代码位置
• 测试结果（通过/失败）
• 若失败，说明原因和尝试的修复方案

注意事项

• 正确性验证：测试用例必须能验证结果的正确性，而不仅仅验证函数能运行
• 测试框架选择：根据需求适配（pytest/unittest/其他）
• 版本号递增：每次成功修复后版本号+1
• 日志记录：所有测试结果和错误信息都要记录到log目录