ww-w-ai-bkit
A Claude Code plugin that checks AI-generated code against its design specs, with audit logs, auth, and database skills for bkend.ai projects. Useful for developers reviewing code that an AI assistant wrote.
What ai-api-test-skill runs automatically, and when. A hook is a command Claude Code fires at a fixed moment, without you asking for it.
Claude Code
$ npx -y skills add buer2233/ai-api-test-skill --agent claude-code
Cursor
Cursor also runs Claude Code hooks from .claude/settings.json, so copy these hooks there to use them in Cursor.
Where it lives
# -*- coding: utf-8 -*-
# Author: dengwanpeng
"""Claude Code PreToolUse hook:在 AI 调用 Skill 工具触发 api-test-E10 时,
自动执行 tools/preflight_check.py,把输出通过 hookSpecificOutput.additionalContext
注入 AI 上下文,等价于 SKILL.md 里原本的「前置必跑 0」。
Hook I/O 协议(Claude Code 标准):
- stdin: JSON {tool_name, tool_input: {skill, args?}, cwd, session_id, ...}
- stdout: JSON {hookSpecificOutput: {hookEventName, additionalContext}}
- exit 0: 放行;exit 2: 阻断该工具调用
⚠️ 触发范围限制:仅在 AI 主动调 `Skill({skill: "api-test-E10"})` 时生效;
被 `Read` / `@skill_path` / system-reminder 触发的 skill 加载不会进入这里。
"""
import json
import os
import subprocess
import sys
HOOK_DIR = os.path.dirname(os.path.abspath(__file__))
SKILL_ROOT = os.path.dirname(HOOK_DIR)
PREFLIGHT_SCRIPT = os.path.join(SKILL_ROOT, "tools", "preflight_check.py")
for _stream in (sys.stdout, sys.stderr):
try:
_stream.reconfigure(encoding="utf-8")
except (AttributeError, ValueError):
pass
def main() -> int:
try:
payload = json.load(sys.stdin)
except Exception:
return 0
tool_input = payload.get("tool_input") or {}
skill_name = (tool_input.get("skill") or "").strip()
# 只拦截 api-test-E10,避免影响其他 Skill。
should_run = skill_name == "api-test-E10"
if not should_run:
return 0
cwd = payload.get("cwd") or os.getcwd()
env = os.environ.copy()
env["PYTHONIOENCODING"] = "utf-8"
try:
result = subprocess.run(
[sys.executable, PREFLIGHT_SCRIPT],
cwd=cwd,
text=True,
encoding="utf-8",
env=env,
capture_output=True,
timeout=120,
)
except Exception as e:
# preflight 自身崩溃(脚本路径错 / 超时 / Python 缺失),不要阻断 skill
err_ctx = f"[preflight_hook] 执行 preflight_check.py 失败: {e}"
print(json.dumps({
"hookSpecificOutput": {
"hookEventName": "PreToolUse",
"additionalContext": err_ctx,
}
}, ensure_ascii=False))
return 0
output_blob = (result.stdout or "") + (result.stderr or "")
context_text = (
"[preflight_check] 入口前置检查结果(由 hook 自动执行,等价于 SKILL.md 原前置必跑 0):\n"
+ output_blob.strip()
)
print(json.dumps({
"hookSpecificOutput": {
"hookEventName": "PreToolUse",
"additionalContext": context_text,
}
}, ensure_ascii=False))
# preflight 失败只把诊断注入上下文,不阻断 Skill 调用。
if result.returncode != 0:
return 0
return 0
if __name__ == "__main__":
sys.exit(main())
Read the script before you install anything that runs on your machine. This is the one part of a plugin that acts without being asked.
面向 Python + pytest + requests 接口自动化项目的 AI Skill:让 Codex / Claude Code 按固定门禁、固定流程、固定 pytest 闭环来新增和维护接口自动化用例,流程图参考:flow_chart/flow.md。 它不是一个普通的提示词模板,而是一套可随仓库分发的工程化 Skill:会先校验任务信息,再按“抓包 / 参考用例 / cURL / Java Controller / pytest 报错”选择路径,最后把接口方法、接口用例和测试验证收敛到同一套规范里。
A Claude Code plugin that checks AI-generated code against its design specs, with audit logs, auth, and database skills for bkend.ai projects. Useful for developers reviewing code that an AI assistant wrote.
Catches accessibility problems while an AI coding assistant writes web interface code, so WCAG accessibility issues are caught before they ship. It also compares audits across commits to flag regressions and drafts an accessibility statement from the findings.
A methodology kit for Claude Code that drives feature work through acceptance test driven development, with spec-first artifacts, architecture checks, and a CRAP analyzer for risk-based refactor and test plans on changed code.
Next-generation UI automation harness powered by Python & Playwright. Chat with AI Agents to seamlessly generate, architect, and execute enterprise-grade UI test code.
Runs a disciplined, test-first workflow for building features and fixing bugs in Claude Code. It brainstorms and plans before coding, reviews finished work against the plan, and guides Playwright end-to-end tests for Next.js and React apps, so you stop skipping tests and debugging by guesswork.