Agent Skills › AgentsMesh/AgentsMesh › e2e

e2e

GitHub

AgentsMesh端到端测试技能，自动选择并运行Web、Desktop、MCP或iOS的E2E套件。负责开发环境搭建、浏览器级验证及故障排查，确保变更行为得到真实栈覆盖。

agents/skills/e2e/SKILL.md AgentsMesh/AgentsMesh

触发场景

需要执行端到端测试诊断E2E测试失败验证跨服务工作流

安装

npx skills add AgentsMesh/AgentsMesh --skill e2e -g -y

更多选项

非标准路径

npx skills add https://github.com/AgentsMesh/AgentsMesh/tree/main/agents/skills/e2e -g -y

不安装直接使用

npx skills use AgentsMesh/AgentsMesh@e2e

指定 Agent (Claude Code)

npx skills add AgentsMesh/AgentsMesh --skill e2e -a claude-code -g -y

安装 repo 全部 skill

npx skills add AgentsMesh/AgentsMesh --all -g -y

预览 repo 内 skill

npx skills add AgentsMesh/AgentsMesh --list

SKILL.md

Frontmatter

{
    "name": "e2e",
    "description": "Selects and runs the appropriate AgentsMesh end-to-end suite for Web, Desktop, MCP, or iOS, including worktree-specific environment setup and browser-level verification. Use when a change needs E2E coverage, a user asks to execute or diagnose an E2E test, or a cross-service workflow must be verified against the real development stack."
}

AgentsMesh E2E

Run the smallest real end-to-end test that proves the requested behavior, then expand coverage in proportion to the change.

Invariants

Use the current worktree's generated deploy/dev/.env; never assume fixed ports or container names.
A UI requirement must be verified by a browser or app UI test. Do not replace it with a raw API request.
Start the development stack only when the selected suite needs it. Record whether this workflow started the stack before deciding whether to clean it.
Authentication uses Connect-RPC. Never restore the removed REST login route; read references/connect-rpc.md when testing auth or proxy routing.
Do not claim a suite passed unless its command completed successfully.
Preserve test artifacts and relevant service logs when a failure occurs.

Workflow

Inspect the requested behavior and changed files.
Select the suite from references/test-matrix.md. Read the selected target's BUILD file, config, and nearby tests before running it.
Reuse a healthy existing stack. Otherwise start //deploy/dev:up for Web flows or //deploy/dev:backend_only when frontends are not required.
Load deploy/dev/.env with automatic export enabled so Bazel can forward the worktree-specific values:
```
set -a
source deploy/dev/.env
set +a
```
Run the narrowest relevant spec or test filter first. If it passes, run the owning target when the blast radius warrants it.
For UI behavior, inspect the rendered state and interaction result through the suite's browser/app assertions. Capture a screenshot or trace when the visual result is central to the bug.
On failure, report the failing assertion and inspect the applicable service logs and test artifacts listed in the test matrix.
Clean the stack with bazel run //deploy/dev:clean only when this workflow started it and no concurrent work depends on it.

Result

Report the exact target and filter, actual worktree ports used, pass/fail status, failing evidence when applicable, and any coverage that could not be run with the reason.