AgentSkillsCN

Testing and CI

运行测试、编写新测试、维护 CI 稳定性的操作指南。

SKILL.md
--- frontmatter
name: Testing and CI
description: Instructions for running tests, writing new tests, and maintaining CI stability.

Testing and CI Guidelines

Core Principles

  1. •Strict Containment: Tests MUST NOT make external API calls (LLM, YouTube, Twitter).
  2. •Fail Fast: Tests should fail immediately if configuration is wrong.

Environment Variables

The test runner (scripts/tasks.ts test) enforces these variables. ALways use them when running tests manually or verifying code.

  • •SKIP_LLM=true: Mocks LLM calls. Agents will try to load output.yaml from their run store instead of calling Gemini.
  • •DRY_RUN=true: Prevents side effects (uploads, posts).

Running Tests

Always use the project task runner:

bash
npm test
# OR
npx tsx scripts/tasks.ts test

Writing New Tests

  1. •Location: Place new tests in tests/.
  2. •Setup: Import tests/setup.js at the top of the file.
  3. •Naming: *.test.ts.
  4. •Mocking:
    • •If testing an Agent, seed the AssetStore with expected "outputs" if you expect SKIP_LLM to work.
    • •Use AssetStore to scaffold a temporary run directory.

CI Configuration

  • •Workflow: .github/workflows/ci.yml
  • •The CI runs npm test.
  • •Dependencies: ensure all dev tools (like fs-extra) are in devDependencies.