Testing Integration

CodeBuddy can run your project’s tests, parse the results, and fix failing tests — all through natural language. It auto-detects your test framework and knows how to filter, run, and interpret results for 6 frameworks.

Supported frameworks

Priority	Framework	Detected via	Command
1	Vitest	package.json scripts/devDeps	npx vitest run
2	Jest	package.json scripts/devDeps	npx jest
3	Mocha	package.json scripts/devDeps	npx mocha / npm test --
4	Pytest	pytest.ini, pyproject.toml, setup.py	python -m pytest
5	Go test	go.mod	go test ./...
6	Cargo test	Cargo.toml	cargo test
fallback	npm test	package.json with non-default test script	npm test --

Detection runs in priority order — the first match wins. You can also override detection entirely with a custom command.

Test name filtering

Each framework supports running a specific test by name:

Framework	Flag
Jest	--testNamePattern
Vitest	-t
Mocha	--grep
Pytest	-k
Go	-run
Cargo	-- (pass-through)

How it works

The run_tests tool

In Agent mode, CodeBuddy uses the run_tests tool to execute tests:

run_tests({ testPath?: "src/auth/", testName?: "should validate token" })

The tool:

1. Detects the framework (or uses your custom command)
2. Spawns the test process with shell: false for security
3. Parses the output into structured results
4. Formats a markdown summary for the agent

Test results

{
  framework: "jest",
  command: "npx jest src/auth/",
  passed: 42,
  failed: 3,
  skipped: 1,
  total: 46,
  duration: "4.2s",
  success: false,
  failures: [
    {
      testName: "should validate expired token",
      file: "src/auth/token.test.ts",
      message: "Expected: false, Received: true",
      expected: "false",
      actual: "true"
    }
  ]
}

The agent receives a formatted summary with pass/fail counts, up to 10 failure details, and the last 2,000 characters of raw output.

Test-fix loops

The tester subagent specializes in test-driven workflows:

1. Analyze — read the code under test
2. Identify — determine what test cases are needed
3. Write — create or update test files
4. Run — execute tests via run_tests
5. Fix — if tests fail, fix the code and re-run

This subagent is invoked automatically during collaboration playbooks:

Workflow	Subagent chain
New feature	Architect → Reviewer → Developer → Tester → Reviewer
Bug fix	Debugger → Developer → Tester
Refactor	Code Analyzer → Architect → File Organizer → Developer → Tester

The tester has access to: run_tests, edit_file, manage_terminal, ripgrep_search, get_diagnostics, list_files, search_vector_db, git, browser, and MCP tools.

Security

Test execution uses strict security measures:

• shell: false — arguments are passed as an array, preventing shell injection
• Argument sanitization — all arguments are validated against shell metacharacters (;|&$(){}!<>) with a 500-character limit
• Environment — forced to FORCE_COLOR=0, CI=true for clean, parseable output
• Custom command parsing — respects single/double quotes properly

Settings

Setting	Type	Default	Description
codebuddy.testCommand	string	""	Custom test command. Overrides auto-detection entirely
codebuddy.testTimeout	number	120000	Test execution timeout in ms (5,000–600,000)

Example prompts

• “Run the tests for the auth module”
• “Run only the test named ‘should validate token’”
• “Write tests for src/utils/parser.ts and run them”
• “Fix the failing tests and make sure they all pass”
• “Add unit tests for all exported functions in this file”