Testing Integration

Detects your test framework, runs tests, parses results, and fixes failing tests via natural-language requests.

Supported frameworks

Detection runs in priority order — first match wins. Or override entirely with codebuddy.testCommand.

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 a non-default test script	npm test --

Test-name filtering per framework:

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

The run_tests tool

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

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

Result shape:

{
  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"
    }
  ]
}

Agent receives a formatted summary with pass/fail counts, up to 10 failure details, and the last 2000 chars of raw output.

Test-fix loops (tester subagent)

Test-driven workflow specialist:

1. Analyze — read code under test.
2. Identify — determine needed test cases.
3. Write — create/update test files.
4. Run — via run_tests.
5. Fix — patch failing code and re-run.

Invoked automatically in collaboration playbooks:

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

Tester has access to run_tests, edit_file, manage_terminal, ripgrep_search, get_diagnostics, list_files, search_vector_db, git, browser, plus MCP tools.

Security

• shell: false — args passed as array, no shell injection.
• Arg sanitization — validated against shell metacharacters (;|&$(){}!<>), 500-char limit each.
• Env — forced to FORCE_COLOR=0, CI=true for clean parseable output.
• Custom command parsing — respects single + double quotes properly.

Settings

Setting	Default	Purpose
codebuddy.testCommand	""	Custom test command — overrides auto-detection
codebuddy.testTimeout	120000	Execution timeout (ms, 5000–600 000)

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”

Related

• Subagents — tester role
• Deep terminal — the terminal path tests run through