SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsdoc-todo-log-loop

doc-todo-log-loop

General & other

View repository

Updated 10-15 17:31

About this skill

Skill: doc-todo-log-loop

1. Overview

A document-driven, log-recorded human-agent collaborative development workflow. The user controls the development pace, and the Agent is responsible for document writing, task decomposition, development execution, and logging.

2. Concept Definitions

The following are the default paths and naming conventions. When the project charter has special provisions, the project charter takes precedence.

  • 文档目录: Defaults to the project root directory backlog/, storing documents such as requirements, design, and logs
  • TODO 文件: Defaults to the project root directory TODO.md
  • 测试目录: Defaults to the project root directory tests/
  • 文档命名: YYYY-MM-DD-HH-mm-{类别}-{标题}.md
  • 测试用例集命名: tests/{YYYY-MM-DD-HH-mm}-testsuite/
  • 测试用例命名: part{序号}-{模块}/case{序号}-{简述}.md, within the test case set

3. Document Categories and Writing Style

In addition to the workflow described later, the Agent may write documents at any time if the user requests it. In document naming, {类别} takes values according to the following categories:

  • 开发日志: Records of the development process, decisions, problems, and solutions.
  • 需求: The functions or goals the user wants to achieve. Contains only the requirements themselves, without implementation details.
  • 设计: Advance analysis of tasks about to be implemented. Subcategories:系统设计、架构设计、交互设计、需求设计。
  • 规范: Defines broadly applicable rules, processes, or standards. Subcategories:架构规范、代码规范、流程规范。
  • 说明: Usage instructions for completed technical entities. Subcategories:接口说明、模块说明。
  • 调研: Research and comparative analysis of external technologies or materials. Search broadly, record sources, and write out details rather than over-summarizing.
  • 参考: Raw materials excerpted from external sources. The difference from research is that it focuses on verbatim quotation rather than active analysis.

The main readers of the documents are future Agents and developers. All documents, process logs, and explanations remain concise, restrained, calm, free of adjectives, and free of metaphors. No lengthy descriptions are needed. Record human decisions, problems, and correction plans. Providing the minimum explanation needed for retrieval and understanding is sufficient.

When the user instructs or the process requires internet research: search broadly for relevant materials; each time a piece of material is found, record it as a separate research document, named according to the naming conventions; the document must record the source and write out the details of the plans, viewpoints, and methods, without over-summarizing.

4. Main Work Cycle

The main workflow defined by this Skill is executed alternately by the user and the Agent, following these steps:

Step 1: Background Description → Document Writing

  • Trigger: The user proposes a functional goal or problem background.
  • Agent Actions:
    1. Communicate with the user to understand the background, goals, and constraints.
    2. Write the requirements description document (naming:YYYY-MM-DD-HH-mm-需求-{简述}.md）。
    3. If there is a Plan file already accepted in Plan Mode, move it to the document directory and name it appropriately.
    4. If the project includes tests, under tests/ prepare the corresponding test case set (see the "Test Case Management" section for the structure).

Step 2: Requirements Description → TODO Decomposition

  • Trigger: The user proposes specific requirements based on documents or directly.
  • Agent Actions:
    1. Break down the requirements into specific, atomic to-do items, and update them to TODO.md。
    2. Each item is associated with the corresponding test case in tests/ .

Step 3: Task Assignment

  • Trigger: The user selects an item from TODO.md and explicitly instructs execution.
  • Agent Actions: Confirm the instruction, understand the necessary documents, understand the testing requirements, then proceed to development and verification.

Step 4: Development and Verification

  • Trigger: The user issues a development instruction.
  • Agent Actions:
    1. Execute the development task.
    2. After completing each feature point, verify it item by item according to the associated test cases (the test cases are a manual checklist; the Agent operates step by step and records the results).
    3. When reporting to the user, the verification results (pass/fail/to be observed) must be included.
    4. The user performs final confirmation.
  • Constraints: When verification has not passed, it is forbidden to claim the task is complete.

Step 5: Development Log Recording

  • Trigger: After the TODO items are confirmed complete by the user.
  • Agent Actions:
    1. Write the development log (naming:YYYY-MM-DD-HH-mm-开发日志-{标题}.md), contents include:
      • Which feature points were implemented
      • Main modifications to the code or project structure
      • Problems encountered and how they were fixed (including both user and Agent)
      • Suggestions for next steps
    2. If testing was performed, the log includes the verification results: which test cases were executed, the corresponding test case set version, and the verification conclusion.
    3. After the log is complete, you may consult TODO.md and suggest the next task to the user, but do not start it on your own initiative.

Step 6: Version Control

  • Trigger: The development log has been written.
  • User Actions: Review the changes and perform the git commit. The Agent is not responsible for this step.

5. Test Case Management

Test cases are a manual checklist in Markdown format, stored in the tests/ directory, versioned by timestamp directories:

tests/YYYY-MM-DD-HH-mm-testsuite/
├── readme.md                         # 概述、环境要求、执行方法
├── part1-{模块名}/
│   ├── case1-{简述}.md
│   └── case2-{简述}.md
└── part2-{模块名}/

Each test case document contains:

  • ID & Title: Unique identifier
  • Preconditions: The initial state required to execute the test
  • Input/Operations: The specific execution steps or data
  • Expected Results: Clear success criteria
  • Actual Results: (Optional) Observations from this run; may also be recorded in the development log

First Creation

In Step 1, directly create a new timestamp directory and write the test cases.

Version Update

  1. Copy the current latest test case set directory in full to a new timestamp directory, adding the -EDITING suffix.
  2. Add, delete, or modify in the new directory until the user confirms.
  3. Remove the -EDITING suffix.
  4. Reference the new version in TODO.md and the development log.