agentcheckup 0.4.1__tar.gz → 0.4.3__tar.gz

{agentcheckup-0.4.1 → agentcheckup-0.4.3}/.github/ISSUE_TEMPLATE/config.yml

@@ -1,8 +1,8 @@
 blank_issues_enabled: false
 contact_links:
   - name: Question or idea
-    url: https://github.com/wanjiaqing/agentcheckup/discussions
+    url: https://github.com/jiaqinq/agentcheckup/discussions
     about: For open-ended discussion, use Discussions instead of an issue.
   - name: Security vulnerability
-    url: https://github.com/wanjiaqing/agentcheckup/security/advisories/new
+    url: https://github.com/jiaqinq/agentcheckup/security/advisories/new
     about: Please report security issues privately, not in a public issue.

{agentcheckup-0.4.1 → agentcheckup-0.4.3}/.github/workflows/publish.yml

@@ -7,7 +7,7 @@ name: Publish
 # publisher"):
 #   https://pypi.org/manage/account/publishing/
 #     PyPI Project Name : agentcheckup
-#     Owner             : wanjiaqing
+#     Owner             : jiaqinq
 #     Repository name   : agentcheckup
 #     Workflow name     : publish.yml
 #     Environment       : pypi

{agentcheckup-0.4.1 → agentcheckup-0.4.3}/CHANGELOG.md

@@ -9,6 +9,24 @@ new ids, and removing a rule is a breaking change.
 
 ## [Unreleased]
 
+## [0.4.2] - 2026-11-07
+
+Three claims the tool made that it had not earned. All were found by an external
+review of the README and the report text, not by a bug report.
+
+### Fixed
+
+- **`AGT201` no longer asserts what the client does.** The hint said "the client
+  keeps one and the other is unreachable". No client was ever tested; the
+  behaviour differs between clients and some namespace the tools instead. The
+  message now says the call can reach either one and that clients resolve it
+  differently, which is true without a test.
+- **README no longer claims "No network calls, ever."** agentcheckup makes none
+  of its own, but `audit` starts every server in your config, and those processes
+  can do whatever they were written to do. The claim now names that boundary and
+  points at `--no-spawn`.
+- **README said 27 config locations; there are 26.** Both languages.
+
 ## [0.4.1] - 2026-11-07
 
 Both fixes came out of an external fact-check of the project, not from a bug

{agentcheckup-0.4.1 → agentcheckup-0.4.3}/CONTRIBUTING.md

@@ -6,7 +6,7 @@ your first one takes about ten minutes.
 ## Setup
 
 ```bash
-git clone https://github.com/wanjiaqing/agentcheckup
+git clone https://github.com/jiaqinq/agentcheckup
 cd agentcheckup
 python -m venv .venv
 # Windows: .venv\Scripts\activate

{agentcheckup-0.4.1 → agentcheckup-0.4.3}/LICENSE

@@ -1,6 +1,6 @@
 MIT License
 
-Copyright (c) 2026 wanjiaqing
+Copyright (c) 2026 jiaqinq
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

{agentcheckup-0.4.1 → agentcheckup-0.4.3}/PKG-INFO

@@ -1,11 +1,11 @@
 Metadata-Version: 2.5
 Name: agentcheckup
-Version: 0.4.1
+Version: 0.4.3
 Summary: Diagnose your agent setup: tool conflicts, risky config, context budget.
-Project-URL: Homepage, https://github.com/wanjiaqing/agentcheckup
-Project-URL: Issues, https://github.com/wanjiaqing/agentcheckup/issues
-Project-URL: Changelog, https://github.com/wanjiaqing/agentcheckup/blob/main/CHANGELOG.md
-Author: wanjiaqing
+Project-URL: Homepage, https://github.com/jiaqinq/agentcheckup
+Project-URL: Issues, https://github.com/jiaqinq/agentcheckup/issues
+Project-URL: Changelog, https://github.com/jiaqinq/agentcheckup/blob/main/CHANGELOG.md
+Author: jiaqinq
 License: MIT
 License-File: LICENSE
 Keywords: agents,ai,cli,linter,mcp,model-context-protocol,security
@@ -30,7 +30,7 @@ Description-Content-Type: text/markdown
 
 > Diagnose your agent setup before your agent does.
 
-[![CI](https://github.com/wanjiaqing/agentcheckup/actions/workflows/ci.yml/badge.svg)](https://github.com/wanjiaqing/agentcheckup/actions/workflows/ci.yml)
+[![CI](https://github.com/jiaqinq/agentcheckup/actions/workflows/ci.yml/badge.svg)](https://github.com/jiaqinq/agentcheckup/actions/workflows/ci.yml)
 [![PyPI](https://img.shields.io/pypi/v/agentcheckup.svg)](https://pypi.org/project/agentcheckup/)
 [![Python](https://img.shields.io/pypi/pyversions/agentcheckup.svg)](https://pypi.org/project/agentcheckup/)
 [![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)
@@ -114,7 +114,7 @@ no network calls. It reads JSON, starts processes, prints text.
 agentcheckup
 ```
 
-Finds your configs automatically — 27 known locations across Claude Desktop,
+Finds your configs automatically — 26 known locations across Claude Desktop,
 Claude Code (including its per-project `mcpServers`), Cursor, VS Code, Cline,
 Roo Code, Kilo Code, Windsurf, Zed, Codex CLI and Continue, plus project-local
 `.mcp.json` / `.cursor/mcp.json` / `.vscode/mcp.json` — starts each server,
@@ -149,7 +149,7 @@ The fastest way to keep a manifest from regressing, and the version you pin is
 the version you run — the action installs from its own checkout.
 
 ```yaml
-- uses: wanjiaqing/agentcheckup@v0.4.1
+- uses: jiaqinq/agentcheckup@v0.4.3
   with:
     config: .mcp.json
     fail-on: warning
@@ -193,6 +193,11 @@ Each tool pair across servers is scored on two signals:
 - **name overlap** — normalised token Jaccard (`read_files` ≡ `read_file`)
 - **description overlap** — token Jaccard over stopword-stripped, plural-stemmed words
 
+Chinese is tokenised as character bigrams (`读取文件` → `读取`, `取文`, `文件`) rather
+than by a segmenter, because a segmenter is a dependency. It is cruder than a real
+word split, but without it every Chinese description produced an empty token set
+and two Chinese tools could never be compared at all.
+
 A pair is flagged when the descriptions overlap ≥ 65% **and** the names share a
 head noun, or when the blended score clears 0.55. The shared-noun requirement is
 what keeps `search_repositories` and `search_documents` apart — without it the
@@ -238,8 +243,9 @@ rule. A report nobody can read is a report nobody acts on.
   to ignore the output.
 - **Degrade, don't die.** A server that will not start is one finding, not an
   aborted run.
-- **No network calls, ever.** Your config never leaves the machine. This is the
-  one thing a commercial scanner structurally cannot promise.
+- **No network calls of its own.** No telemetry, no API, no upload; your config
+  never leaves the machine. It does *start* the servers you configured, and those
+  can do whatever they were written to do — `--no-spawn` skips that entirely.
 - **Heuristics are labelled as heuristics.** `info` means "a human should
   decide". The score is a conversation starter, not a certification.
 
@@ -250,6 +256,10 @@ rule. A report nobody can read is a report nobody acts on.
 - Not an MCP protocol validator. It checks quality and safety, not conformance —
   the official `modelcontextprotocol/conformance` suite does that.
 
+## Further reading
+
+- [多个 MCP Server 一起使用时，如何检测工具冲突？](https://juejin.cn/post/7704995965740580904) — the design story, in Chinese: why the scope changed from a single-server linter to cross-server conflicts, and how the scoring was rebuilt after it graded the official servers 0/100.
+
 ## Contributing
 
 Adding a rule takes about ten minutes and one small file — see

{agentcheckup-0.4.1 → agentcheckup-0.4.3}/README.md

@@ -2,7 +2,7 @@
 
 > Diagnose your agent setup before your agent does.
 
-[![CI](https://github.com/wanjiaqing/agentcheckup/actions/workflows/ci.yml/badge.svg)](https://github.com/wanjiaqing/agentcheckup/actions/workflows/ci.yml)
+[![CI](https://github.com/jiaqinq/agentcheckup/actions/workflows/ci.yml/badge.svg)](https://github.com/jiaqinq/agentcheckup/actions/workflows/ci.yml)
 [![PyPI](https://img.shields.io/pypi/v/agentcheckup.svg)](https://pypi.org/project/agentcheckup/)
 [![Python](https://img.shields.io/pypi/pyversions/agentcheckup.svg)](https://pypi.org/project/agentcheckup/)
 [![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)
@@ -86,7 +86,7 @@ no network calls. It reads JSON, starts processes, prints text.
 agentcheckup
 ```
 
-Finds your configs automatically — 27 known locations across Claude Desktop,
+Finds your configs automatically — 26 known locations across Claude Desktop,
 Claude Code (including its per-project `mcpServers`), Cursor, VS Code, Cline,
 Roo Code, Kilo Code, Windsurf, Zed, Codex CLI and Continue, plus project-local
 `.mcp.json` / `.cursor/mcp.json` / `.vscode/mcp.json` — starts each server,
@@ -121,7 +121,7 @@ The fastest way to keep a manifest from regressing, and the version you pin is
 the version you run — the action installs from its own checkout.
 
 ```yaml
-- uses: wanjiaqing/agentcheckup@v0.4.1
+- uses: jiaqinq/agentcheckup@v0.4.3
   with:
     config: .mcp.json
     fail-on: warning
@@ -165,6 +165,11 @@ Each tool pair across servers is scored on two signals:
 - **name overlap** — normalised token Jaccard (`read_files` ≡ `read_file`)
 - **description overlap** — token Jaccard over stopword-stripped, plural-stemmed words
 
+Chinese is tokenised as character bigrams (`读取文件` → `读取`, `取文`, `文件`) rather
+than by a segmenter, because a segmenter is a dependency. It is cruder than a real
+word split, but without it every Chinese description produced an empty token set
+and two Chinese tools could never be compared at all.
+
 A pair is flagged when the descriptions overlap ≥ 65% **and** the names share a
 head noun, or when the blended score clears 0.55. The shared-noun requirement is
 what keeps `search_repositories` and `search_documents` apart — without it the
@@ -210,8 +215,9 @@ rule. A report nobody can read is a report nobody acts on.
   to ignore the output.
 - **Degrade, don't die.** A server that will not start is one finding, not an
   aborted run.
-- **No network calls, ever.** Your config never leaves the machine. This is the
-  one thing a commercial scanner structurally cannot promise.
+- **No network calls of its own.** No telemetry, no API, no upload; your config
+  never leaves the machine. It does *start* the servers you configured, and those
+  can do whatever they were written to do — `--no-spawn` skips that entirely.
 - **Heuristics are labelled as heuristics.** `info` means "a human should
   decide". The score is a conversation starter, not a certification.
 
@@ -222,6 +228,10 @@ rule. A report nobody can read is a report nobody acts on.
 - Not an MCP protocol validator. It checks quality and safety, not conformance —
   the official `modelcontextprotocol/conformance` suite does that.
 
+## Further reading
+
+- [多个 MCP Server 一起使用时，如何检测工具冲突？](https://juejin.cn/post/7704995965740580904) — the design story, in Chinese: why the scope changed from a single-server linter to cross-server conflicts, and how the scoring was rebuilt after it graded the official servers 0/100.
+
 ## Contributing
 
 Adding a rule takes about ten minutes and one small file — see

{agentcheckup-0.4.1 → agentcheckup-0.4.3}/README.zh-CN.md

@@ -118,6 +118,10 @@ agentcheckup rules                              # 列出全部规则
 - **名字重叠**：归一化后的 token Jaccard（`read_files` ≡ `read_file`）
 - **描述重叠**：去停用词、粗词干化之后的 token Jaccard
 
+中文按**字符二元组**切分（`读取文件` → `读取`、`取文`、`文件`），没有引入分词器——
+分词器就是依赖。这比真正的分词粗，但没有它的时候中文描述会得到空集合，
+两个中文工具根本无法比较。
+
 判定条件：描述重叠 ≥ 65% **且** 名字共享中心词，或加权分 ≥ 0.55。
 共享中心词这个约束是关键——没有它，`search_repositories` 和 `search_documents`
 会因为都含 "search" 被误报，而**充满噪音的报告比没有报告更糟**。
@@ -155,7 +159,8 @@ agentcheckup rules                              # 列出全部规则
 - **规则编号永久不变。** `AGT201` 的含义永远不变。
 - **每条发现都带修法。** 只说「这不好」的规则，训练用户忽略输出。
 - **降级而不是崩溃。** 起不来的 server 是一条发现，不是中断。
-- **永不联网。** 你的配置不会离开本机。这是商业扫描器结构上给不了的承诺。
+- **自己不联网。** 没有遥测、没有 API、不上传任何东西，配置不会离开本机。
+  但它会**启动你配置的 server**，那些进程会做什么取决于它们自己——`--no-spawn` 可完全跳过。
 - **启发式要标明是启发式。** `info` 的意思是「人来判断」。分数是讨论起点，不是认证。
 
 ## 这不是什么
@@ -164,6 +169,10 @@ agentcheckup rules                              # 列出全部规则
 - 不是运行时防护。它不在 agent 会话里运行。
 - 不是 MCP 协议校验器。协议一致性由官方 `modelcontextprotocol/conformance` 负责。
 
+## 相关文章
+
+- [多个 MCP Server 一起使用时，如何检测工具冲突？](https://juejin.cn/post/7704995965740580904) —— 设计复盘：为什么从单 server linter 改成跨 server 冲突检测，以及评分把官方 server 打成 0 分之后怎么重做。
+
 ## 贡献
 
 加一条规则大约十分钟、一个文件，见 [CONTRIBUTING.md](CONTRIBUTING.md)。

{agentcheckup-0.4.1 → agentcheckup-0.4.3}/SECURITY.md

@@ -14,7 +14,7 @@ Nothing leaves your machine.
 ## Reporting a vulnerability
 
 Use GitHub's private advisory flow:
-`https://github.com/wanjiaqing/agentcheckup/security/advisories/new`
+`https://github.com/jiaqinq/agentcheckup/security/advisories/new`
 
 Please do not open a public issue for a vulnerability. Expect an initial response
 within 7 days.

{agentcheckup-0.4.1 → agentcheckup-0.4.3}/action.yml

@@ -1,6 +1,6 @@
 name: agentcheckup
 description: Diagnose your agent setup - cross-server tool conflicts, risky MCP config, context budget.
-author: wanjiaqing
+author: jiaqinq
 
 branding:
   icon: activity

{agentcheckup-0.4.1 → agentcheckup-0.4.3}/docs/github-action.md

@@ -13,7 +13,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
-      - uses: wanjiaqing/agentcheckup@main
+      - uses: jiaqinq/agentcheckup@main
         with:
           config: .mcp.json
           fail-on: warning
@@ -45,7 +45,7 @@ Outputs are published even when the gate fails, so you can gate on score without
 parsing the log:
 
 ```yaml
-      - uses: wanjiaqing/agentcheckup@main
+      - uses: jiaqinq/agentcheckup@main
         id: doctor
         with:
           config: .mcp.json
@@ -64,7 +64,7 @@ permissions:
 
 steps:
   - uses: actions/checkout@v4
-  - uses: wanjiaqing/agentcheckup@main
+  - uses: jiaqinq/agentcheckup@main
     with:
       config: .mcp.json
       sarif-file: agentcheckup.sarif
@@ -87,7 +87,7 @@ for locally. On a fresh runner that usually finds nothing, and agentcheckup exit
 For CI, pass the config explicitly:
 
 ```yaml
-      - uses: wanjiaqing/agentcheckup@main
+      - uses: jiaqinq/agentcheckup@main
         with:
           config: |
             .mcp.json
@@ -99,7 +99,7 @@ For CI, pass the config explicitly:
 Spawning every MCP server takes time. If you only need the config checks:
 
 ```yaml
-      - uses: wanjiaqing/agentcheckup@main
+      - uses: jiaqinq/agentcheckup@main
         with:
           config: .mcp.json
           args: --no-spawn
@@ -113,5 +113,5 @@ are skipped too. It is fast and useful, but it is not the full check.
 `@main` tracks the branch. For reproducible CI, pin a tag or a commit SHA:
 
 ```yaml
-      - uses: wanjiaqing/agentcheckup@v0.3.0
+      - uses: jiaqinq/agentcheckup@v0.3.0
 ```

{agentcheckup-0.4.1 → agentcheckup-0.4.3}/pyproject.toml

@@ -4,12 +4,12 @@ build-backend = "hatchling.build"
 
 [project]
 name = "agentcheckup"
-version = "0.4.1"
+version = "0.4.3"
 description = "Diagnose your agent setup: tool conflicts, risky config, context budget."
 readme = "README.md"
 requires-python = ">=3.10"
 license = { text = "MIT" }
-authors = [{ name = "wanjiaqing" }]
+authors = [{ name = "jiaqinq" }]
 keywords = ["mcp", "model-context-protocol", "agents", "ai", "linter", "security", "cli"]
 classifiers = [
   "Development Status :: 3 - Alpha",
@@ -33,9 +33,9 @@ dev = ["pytest>=8", "ruff>=0.6"]
 agentcheckup = "agentcheckup.cli:main"
 
 [project.urls]
-Homepage = "https://github.com/wanjiaqing/agentcheckup"
-Issues = "https://github.com/wanjiaqing/agentcheckup/issues"
-Changelog = "https://github.com/wanjiaqing/agentcheckup/blob/main/CHANGELOG.md"
+Homepage = "https://github.com/jiaqinq/agentcheckup"
+Issues = "https://github.com/jiaqinq/agentcheckup/issues"
+Changelog = "https://github.com/jiaqinq/agentcheckup/blob/main/CHANGELOG.md"
 
 [tool.hatch.build.targets.wheel]
 packages = ["src/agentcheckup"]

{agentcheckup-0.4.1 → agentcheckup-0.4.3}/src/agentcheckup/conflicts.py

@@ -42,6 +42,10 @@ STOPWORDS = frozenset(
 
 WORD = re.compile(r"[a-z0-9]+")
 
+#: CJK ideographs, extension A included. Matched per character because Chinese
+#: has no word delimiters and a real segmenter would be a dependency.
+CJK = re.compile(r"[\u3400-\u4dbf\u4e00-\u9fff]+")
+
 
 @dataclass(frozen=True)
 class ToolRef:
@@ -94,13 +98,33 @@ def _stem(word: str) -> str:
     return word
 
 
+def _cjk_tokens(text: str) -> set[str]:
+    """Character bigrams over every CJK run.
+
+    Chinese has no spaces, so ``WORD`` returns nothing and two Chinese tools
+    would look like two empty sets — silently, which is the worst outcome. A real
+    segmenter is a dependency; character bigrams are the standard cheap
+    substitute: 读取文件 -> 读取, 取文, 文件.
+
+    A single-character run is kept whole, so 读 still matches 读.
+    """
+    out: set[str] = set()
+    for run in CJK.findall(text):
+        if len(run) == 1:
+            out.add(run)
+        else:
+            out.update(run[i : i + 2] for i in range(len(run) - 1))
+    return out
+
+
 def tokens(text: str) -> set[str]:
-    """Lowercase word set, stopwords removed, crude plural stemming."""
-    return {
+    """Lowercase word set, stopwords removed, crude plural stemming, plus CJK."""
+    latin = {
         _stem(w)
         for w in WORD.findall(text.lower())
         if w not in STOPWORDS and len(w) > 1
     }
+    return latin | _cjk_tokens(text)
 
 
 def jaccard(a: set[str], b: set[str]) -> float:
@@ -111,7 +135,8 @@ def jaccard(a: set[str], b: set[str]) -> float:
 
 
 def name_tokens(name: str) -> set[str]:
-    return {_stem(w) for w in WORD.findall(name.lower()) if len(w) > 1}
+    latin = {_stem(w) for w in WORD.findall(name.lower()) if len(w) > 1}
+    return latin | _cjk_tokens(name)
 
 
 def similarity(a: ToolRef, b: ToolRef) -> float:
@@ -187,7 +212,8 @@ def detect_conflicts(
         name_a = a.tool.name.lower()
         name_b = b.tool.name.lower()
 
-        # Hard collision: identical names, different servers. One is unreachable.
+        # Hard collision: identical names, different servers. What the client
+        # does about it varies, so the message must not promise a specific outcome.
         if name_a == name_b:
             conflicts.append(
                 Conflict(
@@ -201,8 +227,9 @@ def detect_conflicts(
                         f"'{a.server}' and '{b.server}'."
                     ),
                     hint=(
-                        "Hard collision: the client keeps one and the other is "
-                        "unreachable. Rename one, or use a gateway that prefixes."
+                        "Two servers expose the same name and clients resolve that "
+                        "differently, so the same call can reach either one. Rename "
+                        "one, or use a gateway that prefixes."
                     ),
                 )
             )

{agentcheckup-0.4.1 → agentcheckup-0.4.3}/src/agentcheckup/report.py

@@ -10,7 +10,7 @@ from . import rules as _rules
 from .engine import Report
 from .models import Finding, Severity
 
-__version__ = "0.4.1"
+__version__ = "0.4.3"
 
 SARIF_LEVEL = {
     Severity.ERROR: "error",
@@ -195,7 +195,7 @@ def render_text(report: Report, verbose: bool = False) -> str:
         lines.extend(_nudge_lines(nudges))
 
     lines.append("")
-    lines.append("  Rule docs: https://github.com/wanjiaqing/agentcheckup#rules")
+    lines.append("  Rule docs: https://github.com/jiaqinq/agentcheckup#rules")
     return "\n".join(lines)
 
 
@@ -246,7 +246,7 @@ def render_sarif(report: Report) -> str:
                     "driver": {
                         "name": "agentcheckup",
                         "version": __version__,
-                        "informationUri": "https://github.com/wanjiaqing/agentcheckup",
+                        "informationUri": "https://github.com/jiaqinq/agentcheckup",
                         "rules": _rule_metadata(),
                     }
                 },

{agentcheckup-0.4.1 → agentcheckup-0.4.3}/tests/test_conflicts.py

@@ -128,3 +128,40 @@ def test_audit_tokens_sum_across_servers():
     tools = {"a": [tool("alpha_tool")], "b": [tool("beta_tool")]}
     report = lint_audit(servers, tools)
     assert report.tokens > 0
+
+
+# --------------------------------------------------------------------------- #
+# CJK support
+# --------------------------------------------------------------------------- #
+
+
+def test_cjk_runs_become_character_bigrams():
+    assert tokens("读取文件") == {"读取", "取文", "文件"}
+
+
+def test_single_cjk_character_is_kept_whole():
+    assert tokens("读") == {"读"}
+
+
+def test_cjk_and_latin_mix():
+    got = tokens("read 文件")
+    assert "read" in got
+    assert "文件" in got
+
+
+def test_chinese_near_duplicates_are_flagged():
+    """Before CJK support this pair produced two empty token sets and no finding."""
+    a = ref("文件服务", "读取文件", "读取指定路径的文件内容并返回文本")
+    b = ref("文档服务", "读取文档", "读取指定路径的文件内容并返回字符串")
+    assert "AGT203" in {c.rule for c in detect_conflicts([a, b])}
+
+
+def test_chinese_unrelated_tools_are_not_flagged():
+    a = ref("文件服务", "读取文件", "读取指定路径的文件内容并返回文本")
+    b = ref("文件服务", "删除文件", "删除指定路径的文件")
+    assert detect_conflicts([a, b]) == []
+
+
+def test_english_scores_are_unchanged_by_cjk_support():
+    """The CJK path must not touch Latin tokenisation."""
+    assert tokens("Read a file") == {"read", "file"}