AgentSkillsCN

write-tests

创建或更新测试，以反映代码变更。当用户希望为其所做的更改（已暂存或未暂存）或整个功能分支生成/编写测试、确保测试覆盖率，或更新现有测试以匹配代码修改时，请使用此技能。

SKILL.md
--- frontmatter
name: write-tests
description: Create or update tests to reflect code changes. Use when the user wants to generate/write tests for their changes (staged/unstaged) or for an entire feature branch, ensure test coverage, or update existing tests to match code modifications.
argument-hint: "[changes|branch]"
allowed-tools: Bash(npm test *)

Write Tests

Automatically generate or update tests for code changes, ensuring comprehensive test coverage across all branches of logic.

Argument Handling

Check $ARGUMENTS[0] to determine the scope:

  • •changes: Generate tests for all staged and unstaged changes in the working directory
  • •branch: Generate tests for all changes on the current branch compared to the default branch

If $ARGUMENTS[0] is not provided or is neither changes nor branch, respond with:

"Please specify the scope for test generation:

  • •/write-tests changes - Generate tests for staged and unstaged changes
  • •/write-tests branch - Generate tests for all changes on the current branch

Which would you like?"

Workflow

1. Determine Scope and Analyze Changes

If $ARGUMENTS[0] == changes:

Use git commands to identify all modified files:

  • •Run git diff --name-status HEAD to see the names and status of all staged and unstaged changed files
  • •Identify which files have been added, modified, or deleted
  • •Determine the programming languages and frameworks in use

If $ARGUMENTS[0] == branch:

First, identify the branch context:

  • •Determine the current branch name using git branch --show-current
  • •Identify the default branch using git symbolic-ref refs/remotes/origin/HEAD
  • •Verify that the current branch is not the default branch (if it is, inform the user and exit gracefully)

Then, use git commands to identify all changes on the branch:

  • •Run git diff --name-status <default-branch>...HEAD to see the names and status of all files changed by this branch
  • •Identify which files have been added, modified, or deleted since branching
  • •Determine the programming languages and frameworks in use

2. Identify Test Agent Requirements

Based on the file changes:

  • •Check available agents to find specialized test agents (e.g., "backend test agent" for Python, "frontend test agent" for TypeScript/React)
  • •If a specialized test agent exists for the language/framework, note it for later use
  • •Group files by test agent type (backend, frontend, etc.)

3. Determine Test Files

For each modified file:

  • •Identify the corresponding test file (or where it should be created)
  • •Check if the test file exists
  • •Determine if tests need to be created or updated
  • •Group test files by the agent type they require

4. Launch Subagents for Test Generation

For each test file that needs to be created or updated:

  • •Launch a specialized subagent (Task tool) if an appropriate test agent is available, otherwise use a general-purpose agent
  • •Use the instructions from references/subagent-template.md for each subagent
  • •Provide the subagent with the source file changes, test file location, test framework, and test command

IMPORTANT: Launch all subagents in parallel by making multiple Task tool calls in a single message. Do not wait for one to complete before launching the next.

5. Monitor Subagent Completion

  • •Wait for all subagents to complete their work
  • •Review the changes made by each subagent
  • •Check if any subagents reported test failures they couldn't fix after 5 attempts
  • •Note which test files (if any) still have failing tests

6. Run Full Test Suite

Once all subagents have finished:

  • •Identify the project's test command (check package.json, pytest, etc.)
  • •Run the complete test suite for the project
  • •Capture the results

7. Report Results

Provide a summary:

  • •Scope used (changes or branch)
  • •For branch scope: branch name and comparison base
  • •Number of files changed
  • •Number of test files created
  • •Number of test files updated
  • •Final test suite status (pass/fail)
  • •If any test files have persistent failures:
    • •List each test file that still has failures
    • •Include the failure messages from the subagent
    • •Explain that the subagent attempted to fix the failures 5 times but was unsuccessful
    • •Suggest the user manually review these test files

Quality Guidelines

Test Coverage

  • •Tests must cover all code paths, including conditionals, loops, and error handlers
  • •Include positive cases, negative cases, edge cases, and error conditions
  • •Mock external dependencies appropriately
  • •Test both success and failure scenarios

Test Quality

  • •Tests should be clear, focused, and well-named
  • •Each test should verify one specific behavior
  • •Use descriptive assertion messages
  • •Follow the existing test patterns in the codebase

Subagent Instructions

All detailed instructions for subagents are in references/subagent-template.md. This includes:

  • •How to create comprehensive tests
  • •Running tests and fixing failures iteratively (up to 5 attempts)
  • •Reporting results back to the main agent

Error Handling

  • •If $ARGUMENTS[0] is invalid or missing, prompt the user for the correct value
  • •If a subagent fails to complete, report the issue and continue with other agents
  • •If tests cannot be run (missing dependencies, etc.), report the issue clearly
  • •If subagents report persistent failures, surface this information to the user with the test file names and failure details
  • •If no changes are detected, inform the user and exit gracefully

Example Usage

User: "/write-tests changes" User: "/write-tests branch" User: "Generate tests for my changes" → Should invoke with changes scope User: "I need full test coverage before merging this PR" → Should invoke with branch scope