feat: 添加 cann-review 技能-gitcode-skills-AtomGit

已开启

feat: 添加 cann-review 技能 #1

GengChao创建于 3月19日从kobemini/gitcode-skills： feature/cann-review-skill 合入到 stevenaw0/gitcode-skills： main

feat: 添加 cann-review 技能 #1

已开启

GengChao创建于 3月19日

从kobemini/gitcode-skills feature/cann-review-skill合入到stevenaw0/gitcode-skills main

共 6 个文件变更+1546-1

MREADME.md+17-1

	@@ -10,6 +10,9 @@ GitCode Skills 是一组用于支持 Claude Code 操作 GitCode 的技能包。
10	### gitcode-issue-skill	10	### gitcode-issue-skill
11	读取和分析 GitCode issues。支持使用 GitCode API 检索 issue 详情、评论，或分析 cann/ge 仓库中报告的问题。支持认证 API 访问和 markdown 格式化。	11	读取和分析 GitCode issues。支持使用 GitCode API 检索 issue 详情、评论，或分析 cann/ge 仓库中报告的问题。支持认证 API 访问和 markdown 格式化。
12		12
		13	+### cann-review-skill
		14	+审查 GitCode Pull Request，专门用于 CANN 及相关仓库的代码检视。支持快速审查、深度代码审查、行内评论和自动+1。
		15	+
13	## 安装与卸载	16	## 安装与卸载
14	### cluade code	17	### cluade code
15	### 安装	18	### 安装
	@@ -117,6 +120,19 @@ opencode:
117	gitcode-issue 读取https://gitcode.com/cann/ge/issues/153	120	gitcode-issue 读取https://gitcode.com/cann/ge/issues/153
118	```	121	```
119		122
		123	+### 使用 cann-review
		124	+
		125	+当您需要审查 GitCode PR 时：
		126	+
		127	+```bash
		128	+claude code:
		129	+/gitcode-skills:cann-review 审查 https://gitcode.com/cann/metadef/pull/158
		130	+
		131	+opencode:
		132	+/skills
		133	+cann-review 审查 https://gitcode.com/cann/metadef/pull/158
		134	+```
		135	+
120	## 创建新gitcode技能	136	## 创建新gitcode技能
121		137
122	所有新技能必须使用随附的 skill-creator 创建：	138	所有新技能必须使用随附的 skill-creator 创建：
	@@ -168,7 +184,7 @@ gitcode-issue 读取https://gitcode.com/cann/ge/issues/153
168	路线图（欢迎贡献！）：	184	路线图（欢迎贡献！）：
169	- [x] gitcode-issue -拉取issue	185	- [x] gitcode-issue -拉取issue
170	- [x] gitcode-pr - 提交pr，自动按格式填充pr描述	186	- [x] gitcode-pr - 提交pr，自动按格式填充pr描述
171	-- [ ] gitcode-review- - 提交检视意见，读取检视意见	187	+- [x] cann-review - 审查pr，支持深度代码审查和行内评论
172	- [ ] gitcode-ci - 触发ci流水线，获取结果，分析失败原因	188	- [ ] gitcode-ci - 触发ci流水线，获取结果，分析失败原因
173		189
174	## 支持	190	## 支持

Askills/cann-review/README.md+117-0

@@ -0,0 +1,117 @@
	1	+# CANN Review Skill
	2	+
	3	+GitCode Pull Request 代码审查 Skill，用于 CANN 及相关仓库的 PR 审查。
	4	+
	5	+## 功能
	6	+
	7	+- 📊 快速审查: 自动获取 PR 元数据、文件变更、代码差异
	8	+- 🔍 深度代码审查: 逐行检查代码正确性、安全性、内存安全、代码风格
	9	+- 💬 行内评论: 支持在指定代码行添加评论
	10	+- ✅ 自动 LGTM: 低风险 PR 可自动批准
	11	+
	12	+## 安装
	13	+
	14	+### 1. 复制 Skill 目录
	15	+
	16	+将整个 `cann-review` 目录复制到你的 Claude skills 目录：
	17	+
	18	+```bash
	19	+cp -r cann-review ~/.claude/skills/
	20	+```
	21	+
	22	+### 2. 配置 GitCode API Token
	23	+
	24	+创建配置文件 `~/.cann-review/config`：
	25	+
	26	+```bash
	27	+mkdir -p ~/.cann-review
	28	+cat > ~/.cann-review/config << EOF
	29	+GITCODE_TOKEN=your_token_here
	30	+EOF
	31	+```
	32	+
	33	+或者设置环境变量：
	34	+
	35	+```bash
	36	+export GITCODE_TOKEN=your_token_here
	37	+```
	38	+
	39	+## 依赖
	40	+
	41	+- Python 3.7+
	42	+- 仅使用 Python 标准库，无需安装任何第三方包
	43	+
	44	+## 使用方法
	45	+
	46	+### 快速审查
	47	+
	48	+```
	49	+检视 https://gitcode.com/cann/ge/pull/1130
	50	+```
	51	+
	52	+### 深度代码审查
	53	+
	54	+```
	55	+详细检视代码 https://gitcode.com/cann/ge/pull/1130
	56	+```
	57	+
	58	+### 直接 LGTM
	59	+
	60	+```
	61	+直接加 lgtm https://gitcode.com/cann/ge/pull/1130
	62	+```
	63	+
	64	+## 代码审查维度
	65	+
	66	+Skill 会对以下维度进行自动检查：
	67	+
	68	+\| 维度 \| 检查内容 \|
	69	+\|------\|----------\|
	70	+\| 正确性 \| 逻辑错误、边界检查、错误处理 \|
	71	+\| 安全性 \| 缓冲区溢出、注入攻击、输入验证 \|
	72	+\| 内存安全 \| 内存泄漏、双重释放、悬空指针 \|
	73	+\| 代码风格 \| 命名规范、代码重复、魔法数 \|
	74	+\| 测试覆盖 \| 单元测试、边界值、错误路径 \|
	75	+\| 代码复用 \| 抽象设计、DRY 原则 \|
	76	+
	77	+## 自动问题检测
	78	+
	79	+当审查代码发现以下问题时，Skill 会主动询问是否添加行内评论：
	80	+
	81	+- ⚠️ 内存泄漏
	82	+- ⚠️ 安全漏洞
	83	+- ⚠️ 使用过时 API
	84	+- ⚠️ 关键逻辑错误
	85	+- ⚠️ 缺少输入验证
	86	+- ⚠️ 线程安全问题
	87	+
	88	+## 目录结构
	89	+
	90	+```
	91	+cann-review/
	92	+├── SKILL.md # Skill 定义文件
	93	+├── README.md # 本文件
	94	+├── agents/
	95	+│ └── openai.yaml # Agent 配置
	96	+└── scripts/
	97	+ ├── gitcode_pr_review.py # PR 审查主脚本
	98	+ └── gitcode_add_line_comment.py # 行内评论脚本
	99	+```
	100	+
	101	+## 支持的 GitCode URL 格式
	102	+
	103	+- `https://gitcode.com/cann/ge/pull/1130`
	104	+- `https://gitcode.com/cann/ge/pulls/1130`
	105	+- `https://gitcode.com/cann/ge/merge_request/1130`
	106	+- `https://gitcode.com/cann/ge/merge_requests/1130`
	107	+
	108	+## 获取 GitCode API Token
	109	+
	110	+1. 访问 https://gitcode.com/
	111	+2. 登录后进入设置
	112	+3. 生成个人访问令牌 (Personal Access Token)
	113	+4. 将 Token 配置到 `~/.cann-review/config`
	114	+
	115	+## 许可
	116	+
	117	+本 Skill 仅供内部使用。

Askills/cann-review/SKILL.md+265-0

@@ -0,0 +1,265 @@
	1	+---
	2	+name: cann-review
	3	+description: Review GitCode pull requests for CANN and related repositories. Use when the user asks to review a GitCode PR, provides a GitCode PR URL, or mentions CANN PR review. Supports GitCode URL variants like /pull/<n>, /pulls/<n>, and /merge_requests/<n>, then fetches PR metadata and changed files for structured review.
	4	+---
	5	+
	6	+# CANN Review
	7	+
	8	+Use this skill when the user wants a review of a GitCode PR, especially for `cann/*` repositories.
	9	+
	10	+## Workflow
	11	+
	12	+### 1. 获取访问令牌（第一步必须）
	13	+
	14	+检查环境变量：
	15	+```bash
	16	+echo $GITCODE_TOKEN
	17	+```
	18	+
	19	+如果不存在，提示用户获取令牌：
	20	+1. 登录 [GitCode](https://gitcode.com)
	21	+2. 点击头像 → 设置 → 访问令牌
	22	+3. 创建新令牌，选择 `read_repository` 权限
	23	+4. 复制令牌，建议保存到 `~/.bashrc`：`export GITCODE_TOKEN="your_token_here"`
	24	+
	25	+### 2. Normalize the PR URL
	26	+
	27	+```bash
	28	+python3 scripts/gitcode_pr_review.py "<pr_url>"
	29	+```
	30	+
	31	+### 3. Review Decision
	32	+
	33	+Default output is a quick review decision:
	34	+- `可直接 LGTM`
	35	+- `建议人工复查`
	36	+- `不建议直接 LGTM`
	37	+
	38	+### 4. 如果 `GITCODE_TOKEN` 已设置，脚本还会获取：
	39	+- PR title, state, author, branches, labels
	40	+- changed files
	41	+- patch excerpts, with size limits
	42	+
	43	+4. Accept these GitCode page URL kinds:
	44	+- `/pull/<number>`
	45	+- `/pulls/<number>`
	46	+- `/merge_request/<number>`
	47	+- `/merge_requests/<number>`
	48	+
	49	+5. Treat GitCode's API path as canonical:
	50	+- page URLs may be `/pull/...` or `/merge_requests/...`
	51	+- API URLs should use `/api/v5/repos/<project>/pulls/<number>`
	52	+
	53	+6. Review findings first.
	54	+- Prioritize correctness issues, regressions, risky deletions, build/test impact, and missing validation.
	55	+- Keep summaries brief after findings.
	56	+- Do not claim file contents you did not fetch.
	57	+
	58	+7. Optional publishing actions require explicit flags.
	59	+- `--post-comment` posts the generated review/report back to GitCode
	60	+- `--auto-lgtm` only posts the literal `/lgtm` when the decision is `可直接 LGTM`
	61	+- `--lgtm-only` posts the literal `/lgtm` immediately, without waiting for an automatic low-risk decision
	62	+
	63	+## Commands
	64	+
	65	+Parse only:
	66	+
	67	+```bash
	68	+python3 scripts/gitcode_pr_review.py "https://gitcode.com/cann/metadef/pull/158" --skip-files
	69	+```
	70	+
	71	+Fetch quick review:
	72	+
	73	+```bash
	74	+python3 scripts/gitcode_pr_review.py "https://gitcode.com/cann/metadef/pull/158"
	75	+```
	76	+
	77	+Fetch full report:
	78	+
	79	+```bash
	80	+python3 scripts/gitcode_pr_review.py "https://gitcode.com/cann/metadef/pull/158" --format report
	81	+```
	82	+
	83	+Fetch raw context markdown:
	84	+
	85	+```bash
	86	+python3 scripts/gitcode_pr_review.py "https://gitcode.com/cann/metadef/pull/158" --format context-markdown
	87	+```
	88	+
	89	+Write JSON context to a file:
	90	+
	91	+```bash
	92	+python3 scripts/gitcode_pr_review.py "https://gitcode.com/cann/metadef/pull/158" --format json --output /tmp/pr_context.json
	93	+```
	94	+
	95	+Post review report to GitCode:
	96	+
	97	+```bash
	98	+python3 scripts/gitcode_pr_review.py "https://gitcode.com/cann/metadef/pull/158" --format report --post-comment
	99	+```
	100	+
	101	+Post report and auto `/lgtm` when the change is low risk:
	102	+
	103	+```bash
	104	+python3 scripts/gitcode_pr_review.py "https://gitcode.com/cann/metadef/pull/158" --format report --post-comment --auto-lgtm
	105	+```
	106	+
	107	+Post the literal `/lgtm` immediately:
	108	+
	109	+```bash
	110	+python3 scripts/gitcode_pr_review.py "https://gitcode.com/cann/metadef/pull/158" --lgtm-only
	111	+```
	112	+
	113	+### Line Comment Commands
	114	+
	115	+Add a line comment to specific file and line:
	116	+
	117	+```bash
	118	+python3 scripts/gitcode_add_line_comment.py \
	119	+ "https://gitcode.com/cann/metadef/pull/159" \
	120	+ --file "CMakeLists.txt" \
	121	+ --line 128 \
	122	+ --body "⚠️ 使用过时的CMake变量: 建议改为现代写法"
	123	+```
	124	+
	125	+Dry-run to preview comment:
	126	+
	127	+```bash
	128	+python3 scripts/gitcode_add_line_comment.py \
	129	+ "https://gitcode.com/cann/metadef/pull/159" \
	130	+ --file "CMakeLists.txt" \
	131	+ --line 128 \
	132	+ --body "Comment text" \
	133	+ --dry-run
	134	+```
	135	+
	136	+Post as general PR comment (not line-specific):
	137	+
	138	+```bash
	139	+python3 scripts/gitcode_add_line_comment.py \
	140	+ "https://gitcode.com/cann/metadef/pull/159" \
	141	+ --file "N/A" \
	142	+ --line 1 \
	143	+ --body "General comment" \
	144	+ --general
	145	+```
	146	+
	147	+## Rules
	148	+
	149	+- If the user gives a GitCode PR URL, run the script first and use the default quick review unless raw context is explicitly needed.
	150	+- If the user explicitly says "直接加 lgtm" or equivalent, use `--lgtm-only` so the posted content is the literal `/lgtm`, not plain `LGTM`.
	151	+- If `GITCODE_TOKEN` is missing, explain that only URL parsing succeeded and remote PR data was not fetched.
	152	+- For large PRs, rely on the summary plus selected patch excerpts; call out that the context is truncated.
	153	+- Only post comments back to GitCode if the user explicitly asks for that action.
	154	+
	155	+## Deep Code Review Rules
	156	+
	157	+When user asks for detailed code review ("详细审查代码", "逐行检视", "代码检视"), perform comprehensive analysis:
	158	+
	159	+### 1. Fetch Full Context
	160	+Always use `--format context-markdown` to get complete code diffs.
	161	+
	162	+### 2. Review Dimensions
	163	+
	164	+#### Correctness (正确性)
	165	+- Check for logic errors, off-by-one errors, incorrect comparisons
	166	+- Verify bug fixes actually fix the bug
	167	+- Check for proper error handling and null pointer checks
	168	+
	169	+#### Security (安全性)
	170	+- Buffer overflows, out-of-bounds access
	171	+- SQL injection, command injection possibilities
	172	+- Missing input validation
	173	+- Use of unsafe functions (strcpy, sprintf, etc.)
	174	+- Integer overflow/underflow
	175	+- Uninitialized variables
	176	+
	177	+#### Memory Safety (内存安全)
	178	+- Memory leaks (missing free/delete)
	179	+- Double free, use-after-free
	180	+- Dangling pointers
	181	+- Smart pointer misuse
	182	+
	183	+#### Code Style (代码风格)
	184	+- Naming conventions (snake_case for functions, PascalCase for classes in C++)
	185	+- Code duplication
	186	+- Magic numbers without constants
	187	+- Inconsistent formatting
	188	+- Overly complex functions (>100 lines, deep nesting)
	189	+
	190	+#### Testing Coverage (测试覆盖)
	191	+- Are there unit tests for new functionality?
	192	+- Edge cases covered (null, empty, boundary values)?
	193	+- Error path testing?
	194	+- Negative test cases?
	195	+
	196	+#### Code Reusability (代码复用性)
	197	+- Extract common logic into shared functions?
	198	+- Proper abstractions?
	199	+- DRY principle violations?
	200	+
	201	+### 3. Severity Levels
	202	+
	203	+Critical (阻塞) - Must fix before merge:
	204	+- Memory leaks
	205	+- Security vulnerabilities
	206	+- Crashes/assertion failures
	207	+- Data corruption
	208	+
	209	+High (严重) - Should fix:
	210	+- Logic errors
	211	+- Missing error handling
	212	+- Resource leaks
	213	+- Thread safety issues
	214	+
	215	+Medium (中等) - Consider fixing:
	216	+- Code duplication
	217	+- Poor naming
	218	+- Missing test coverage
	219	+- Deprecated API usage
	220	+
	221	+Low (轻微) - Nice to have:
	222	+- Style inconsistencies
	223	+- Minor optimizations
	224	+- Comment improvements
	225	+
	226	+### 4. Automatic Issue Detection
	227	+
	228	+When reviewing, if you detect any of the following, proactively ask the user if they want to add a line comment:
	229	+
	230	+```
	231	+"发现 [问题类型]，是否添加行内评论到PR？"
	232	+```
	233	+
	234	+Trigger issues for automatic prompting:
	235	+- ⚠️ Memory leaks
	236	+- ⚠️ Security vulnerabilities (overflow, injection, etc.)
	237	+- ⚠️ Use of deprecated APIs
	238	+- ⚠️ Critical logic errors
	239	+- ⚠️ Missing validation on user input
	240	+- ⚠️ Thread safety issues
	241	+- ⚠️ Code that will fail compilation on some platforms
	242	+
	243	+### 5. Interactive Comment Posting
	244	+
	245	+After presenting review findings, if the user confirms they want to add a comment:
	246	+
	247	+```
	248	+"已将审查意见作为行内评论添加到 [文件:行号]"
	249	+```
	250	+
	251	+Use the `gitcode_add_line_comment.py` script:
	252	+
	253	+```bash
	254	+python3 scripts/gitcode_add_line_comment.py \
	255	+ "<pr_url>" \
	256	+ --file "<file_path>" \
	257	+ --line <line_number> \
	258	+ --body "<comment_body>"
	259	+```
	260	+
	261	+Comment body format:
	262	+- Use ⚠️ emoji for warnings
	263	+- Use ✅ for positives
	264	+- Provide code examples in markdown code blocks
	265	+- Keep it concise but actionable

Askills/cann-review/agents/openai.yaml+4-0

@@ -0,0 +1,4 @@
	1	+interface:
	2	+ display_name: "CANN Review"
	3	+ short_description: "Review GitCode CANN PRs with robust URL parsing."
	4	+ default_prompt: "Review this GitCode PR. Normalize the URL first, fetch PR metadata and changed files, then report findings ordered by severity."

Askills/cann-review/scripts/gitcode_add_line_comment.py+223-0

@@ -0,0 +1,223 @@
	1	+#!/usr/bin/env python3
	2	+# -- coding: utf-8 --
	3	+"""
	4	+GitCode PR Line Comment Poster
	5	+
	6	+Add line comments to GitCode pull requests.
	7	+Supports both inline comments on specific lines and general PR comments.
	8	+Uses only Python standard library - no external dependencies.
	9	+"""
	10	+
	11	+import argparse
	12	+import json
	13	+import os
	14	+import sys
	15	+from urllib.parse import quote, urlparse
	16	+from urllib.request import Request, urlopen
	17	+from urllib.error import HTTPError, URLError
	18	+
	19	+
	20	+def parse_pr_url(pr_url: str) -> dict:
	21	+ """Parse GitCode PR URL and extract components."""
	22	+ import re
	23	+
	24	+ PR_PATH_RE = re.compile(
	25	+ r"^/(?P<project>(?:[^/]+/){1,2}[^/]+)/"
	26	+ r"(?P<kind>pull\|pulls\|merge_request\|merge_requests)/"
	27	+ r"(?P<number>[0-9]+)"
	28	+ r"/*$"
	29	+ )
	30	+
	31	+ parsed = urlparse(pr_url)
	32	+ if not parsed.netloc or "gitcode" not in parsed.netloc.lower():
	33	+ raise ValueError("Not a GitCode URL")
	34	+
	35	+ match = PR_PATH_RE.match(parsed.path.rstrip("/"))
	36	+ if not match:
	37	+ raise ValueError(
	38	+ "Unsupported GitCode PR URL path. Accepted path kinds: "
	39	+ "/pull/<n>, /pulls/<n>, /merge_request/<n>, /merge_requests/<n>"
	40	+ )
	41	+
	42	+ project_path = match.group("project").strip("/")
	43	+ project_parts = [part for part in project_path.split("/") if part]
	44	+ if len(project_parts) < 2:
	45	+ raise ValueError("GitCode project path must contain at least namespace/repo")
	46	+
	47	+ number = int(match.group("number"))
	48	+
	49	+ return {
	50	+ "project_path": project_path,
	51	+ "pr_number": number,
	52	+ "canonical_web_url": f"https://gitcode.com/{project_path}/merge_requests/{number}",
	53	+ "api_base": f"https://api.gitcode.com/api/v5/repos/{quote(project_path, safe='/')}",
	54	+ }
	55	+
	56	+
	57	+def get_token() -> str:
	58	+ """Get GitCode API token from environment."""
	59	+ token = os.environ.get("GITCODE_TOKEN")
	60	+ if not token:
	61	+ config_path = os.path.expanduser("~/.cann-review/config")
	62	+ if os.path.exists(config_path):
	63	+ with open(config_path, "r", encoding="utf-8") as f:
	64	+ for line in f:
	65	+ line = line.strip()
	66	+ if line.startswith("GITCODE_TOKEN="):
	67	+ token = line.split("=", 1)[1].strip().strip('"').strip("'")
	68	+ break
	69	+ if not token:
	70	+ raise ValueError(
	71	+ "GITCODE_TOKEN not set. Please export GITCODE_TOKEN environment variable."
	72	+ )
	73	+ return token
	74	+
	75	+
	76	+def api_get(url: str, headers: dict) -> dict:
	77	+ """Perform GET request to API."""
	78	+ req = Request(url, headers=headers)
	79	+ try:
	80	+ with urlopen(req, timeout=30) as resp:
	81	+ return json.loads(resp.read().decode("utf-8"))
	82	+ except HTTPError as e:
	83	+ raise RuntimeError(f"HTTP {e.code}: {e.reason}") from e
	84	+ except URLError as e:
	85	+ raise RuntimeError(f"URL Error: {e.reason}") from e
	86	+
	87	+
	88	+def api_post(url: str, headers: dict, data: dict) -> dict:
	89	+ """Perform POST request to API."""
	90	+ body = json.dumps(data).encode("utf-8")
	91	+ req = Request(url, data=body, headers=headers, method="POST")
	92	+ try:
	93	+ with urlopen(req, timeout=30) as resp:
	94	+ response_data = resp.read().decode("utf-8")
	95	+ return {"status_code": resp.status, "success": resp.status in [200, 201], "response": response_data}
	96	+ except HTTPError as e:
	97	+ response_data = e.read().decode("utf-8") if e.fp else ""
	98	+ return {"status_code": e.code, "success": False, "response": response_data}
	99	+ except URLError as e:
	100	+ return {"status_code": 0, "success": False, "response": str(e.reason)}
	101	+
	102	+
	103	+def get_pr_head_sha(api_base: str, pr_number: int, headers: dict) -> str:
	104	+ """Get the HEAD commit SHA of the PR."""
	105	+ url = f"{api_base}/pulls/{pr_number}"
	106	+ data = api_get(url, headers)
	107	+ return data.get("head", {}).get("sha", "")
	108	+
	109	+
	110	+def post_line_comment(
	111	+ api_base: str,
	112	+ pr_number: int,
	113	+ head_sha: str,
	114	+ file_path: str,
	115	+ line: int,
	116	+ body: str,
	117	+ headers: dict,
	118	+) -> dict:
	119	+ """Post a line comment to a PR."""
	120	+ url = f"{api_base}/pulls/{pr_number}/comments"
	121	+
	122	+ payload = {
	123	+ "body": body,
	124	+ "commit_id": head_sha,
	125	+ "path": file_path,
	126	+ "position": line,
	127	+ }
	128	+
	129	+ return api_post(url, headers, payload)
	130	+
	131	+
	132	+def post_pr_comment(
	133	+ api_base: str,
	134	+ pr_number: int,
	135	+ body: str,
	136	+ headers: dict,
	137	+ reply_to_comment_id: str = None,
	138	+) -> dict:
	139	+ """Post a general comment to a PR."""
	140	+ url = f"{api_base}/pulls/{pr_number}/comments"
	141	+
	142	+ payload = {"body": body}
	143	+ if reply_to_comment_id:
	144	+ # For replying to an existing comment (if supported)
	145	+ payload["in_reply_to"] = reply_to_comment_id
	146	+
	147	+ return api_post(url, headers, payload)
	148	+
	149	+
	150	+def main():
	151	+ parser = argparse.ArgumentParser(
	152	+ description="Add line comments to GitCode pull requests"
	153	+ )
	154	+ parser.add_argument("pr_url", help="GitCode PR URL")
	155	+ parser.add_argument("--file", required=True, help="File path to comment on")
	156	+ parser.add_argument("--line", type=int, required=True, help="Line number to comment on")
	157	+ parser.add_argument("--body", required=True, help="Comment body")
	158	+ parser.add_argument(
	159	+ "--general",
	160	+ action="store_true",
	161	+ help="Post as general PR comment instead of line comment",
	162	+ )
	163	+ parser.add_argument(
	164	+ "--dry-run",
	165	+ action="store_true",
	166	+ help="Print what would be posted without actually posting",
	167	+ )
	168	+
	169	+ args = parser.parse_args()
	170	+
	171	+ try:
	172	+ parsed = parse_pr_url(args.pr_url)
	173	+ token = get_token()
	174	+ headers = {
	175	+ "Authorization": f"Bearer {token}",
	176	+ "Content-Type": "application/json",
	177	+ }
	178	+
	179	+ head_sha = get_pr_head_sha(parsed["api_base"], parsed["pr_number"], headers)
	180	+
	181	+ if args.dry_run:
	182	+ print("[Dry Run] Would post the following comment:")
	183	+ print(f" PR: {parsed['canonical_web_url']}")
	184	+ print(f" File: {args.file}")
	185	+ print(f" Line: {args.line}")
	186	+ print(f" HEAD SHA: {head_sha}")
	187	+ print(f" Body:\n{args.body}")
	188	+ return 0
	189	+
	190	+ if args.general:
	191	+ result = post_pr_comment(
	192	+ parsed["api_base"], parsed["pr_number"], args.body, headers
	193	+ )
	194	+ else:
	195	+ result = post_line_comment(
	196	+ parsed["api_base"],
	197	+ parsed["pr_number"],
	198	+ head_sha,
	199	+ args.file,
	200	+ args.line,
	201	+ args.body,
	202	+ headers,
	203	+ )
	204	+
	205	+ if result["success"]:
	206	+ print(f"✓ Comment posted successfully!")
	207	+ print(f" PR: {parsed['canonical_web_url']}")
	208	+ if not args.general:
	209	+ print(f" File: {args.file}:{args.line}")
	210	+ return 0
	211	+ else:
	212	+ print(f"✗ Failed to post comment")
	213	+ print(f" Status: {result['status_code']}")
	214	+ print(f" Error: {result['response']}")
	215	+ return 1
	216	+
	217	+ except Exception as e:
	218	+ print(f"Error: {e}", file=sys.stderr)
	219	+ return 1
	220	+
	221	+
	222	+if __name__ == "__main__":
	223	+ sys.exit(main())

Askills/cann-review/scripts/gitcode_pr_review.py+920-0

@@ -0,0 +1,920 @@
	1	+#!/usr/bin/env python3
	2	+
	3	+from __future__ import annotations
	4	+
	5	+import argparse
	6	+import collections
	7	+from datetime import datetime
	8	+import json
	9	+import os
	10	+import re
	11	+import sys
	12	+from typing import Any
	13	+from urllib.error import HTTPError, URLError
	14	+from urllib.parse import quote, urlparse
	15	+from urllib.request import Request, urlopen
	16	+
	17	+
	18	+PR_PATH_RE = re.compile(
	19	+ r"^/(?P<project>.+?)/(?P<kind>pull\|pulls\|merge_request\|merge_requests\|merge-requests)/(?P<number>\d+)(?:/.*)?$"
	20	+)
	21	+
	22	+SUPPORTED_KINDS = ("pull", "pulls", "merge_request", "merge_requests", "merge-requests")
	23	+
	24	+
	25	+class GitCodeApiError(RuntimeError):
	26	+ pass
	27	+
	28	+
	29	+def read_token_from_config(config_path: str) -> str:
	30	+ try:
	31	+ with open(config_path, "r", encoding="utf-8") as handle:
	32	+ for raw_line in handle:
	33	+ line = raw_line.strip()
	34	+ if not line or line.startswith("#") or "=" not in line:
	35	+ continue
	36	+ key, value = line.split("=", 1)
	37	+ if key.strip() != "GITCODE_TOKEN":
	38	+ continue
	39	+ return value.strip().strip('"').strip("'")
	40	+ except OSError:
	41	+ return ""
	42	+ return ""
	43	+
	44	+
	45	+def resolve_token(token_env: str) -> tuple[str, str]:
	46	+ env_token = os.environ.get(token_env, "").strip()
	47	+ if env_token:
	48	+ return env_token, f"env:{token_env}"
	49	+
	50	+ script_dir = os.path.dirname(os.path.abspath(__file__))
	51	+ candidate_paths = [
	52	+ os.path.expanduser("~/.cann-review/config"),
	53	+ os.path.normpath(os.path.join(script_dir, "..", "config", "gitcode.conf")),
	54	+ ]
	55	+
	56	+ seen: set[str] = set()
	57	+ for candidate in candidate_paths:
	58	+ normalized = os.path.abspath(candidate)
	59	+ if normalized in seen:
	60	+ continue
	61	+ seen.add(normalized)
	62	+ token = read_token_from_config(normalized)
	63	+ if token:
	64	+ return token, f"config:{normalized}"
	65	+
	66	+ return "", "missing"
	67	+
	68	+
	69	+def parse_pr_url(pr_url: str) -> dict[str, Any]:
	70	+ parsed = urlparse(pr_url.strip())
	71	+ if parsed.scheme not in {"http", "https"}:
	72	+ raise ValueError("PR URL must start with http:// or https://")
	73	+ if parsed.netloc not in {"gitcode.com", "www.gitcode.com"}:
	74	+ raise ValueError("PR URL must point to gitcode.com")
	75	+
	76	+ match = PR_PATH_RE.match(parsed.path.rstrip("/"))
	77	+ if not match:
	78	+ accepted = ", ".join(SUPPORTED_KINDS)
	79	+ raise ValueError(
	80	+ "Unsupported GitCode PR URL path. Accepted path kinds: "
	81	+ f"{accepted}. Example: https://gitcode.com/cann/metadef/pull/158"
	82	+ )
	83	+
	84	+ project_path = match.group("project").strip("/")
	85	+ project_parts = [part for part in project_path.split("/") if part]
	86	+ if len(project_parts) < 2:
	87	+ raise ValueError("GitCode project path must contain at least namespace/repo")
	88	+
	89	+ number = int(match.group("number"))
	90	+ input_kind = match.group("kind")
	91	+ canonical_web_url = f"https://gitcode.com/{project_path}/merge_requests/{number}"
	92	+ api_url = f"https://api.gitcode.com/api/v5/repos/{quote(project_path, safe='/')}/pulls/{number}"
	93	+
	94	+ return {
	95	+ "source_url": pr_url,
	96	+ "project_path": project_path,
	97	+ "namespace_path": "/".join(project_parts[:-1]),
	98	+ "repo_name": project_parts[-1],
	99	+ "pr_number": number,
	100	+ "input_kind": input_kind,
	101	+ "canonical_web_url": canonical_web_url,
	102	+ "api_url": api_url,
	103	+ "files_api_url": f"{api_url}/files",
	104	+ "comments_api_url": f"{api_url}/comments",
	105	+ }
	106	+
	107	+
	108	+def request_json(url: str, token: str, timeout: float, method: str = "GET", data: Any = None) -> Any:
	109	+ headers = {
	110	+ "Accept": "application/json",
	111	+ "Authorization": f"Bearer {token}",
	112	+ }
	113	+ payload = None
	114	+ if data is not None:
	115	+ headers["Content-Type"] = "application/json"
	116	+ payload = json.dumps(data, ensure_ascii=False).encode("utf-8")
	117	+ request = Request(url, headers=headers, method=method, data=payload)
	118	+ try:
	119	+ with urlopen(request, timeout=timeout) as response:
	120	+ return json.loads(response.read().decode("utf-8"))
	121	+ except HTTPError as exc:
	122	+ body = exc.read().decode("utf-8", errors="replace")
	123	+ raise GitCodeApiError(f"GitCode API request failed: HTTP {exc.code} for {url}: {body}") from exc
	124	+ except URLError as exc:
	125	+ raise GitCodeApiError(f"GitCode API request failed for {url}: {exc}") from exc
	126	+
	127	+
	128	+def fetch_json(url: str, token: str, timeout: float) -> Any:
	129	+ return request_json(url, token, timeout, method="GET")
	130	+
	131	+
	132	+def post_comment(url: str, token: str, timeout: float, body: str) -> Any:
	133	+ return request_json(url, token, timeout, method="POST", data={"body": body})
	134	+
	135	+
	136	+def infer_file_status(file_item: dict[str, Any]) -> str:
	137	+ if file_item.get("status"):
	138	+ return str(file_item["status"])
	139	+ if file_item.get("new_file"):
	140	+ return "added"
	141	+ if file_item.get("deleted_file"):
	142	+ return "deleted"
	143	+ if file_item.get("renamed_file"):
	144	+ return "renamed"
	145	+ return "modified"
	146	+
	147	+
	148	+def extract_patch(file_item: dict[str, Any]) -> str:
	149	+ patch = file_item.get("patch")
	150	+ if isinstance(patch, dict):
	151	+ return str(patch.get("diff") or "")
	152	+ if isinstance(patch, str):
	153	+ return patch
	154	+ return ""
	155	+
	156	+
	157	+def clip_text(text: str, limit: int) -> tuple[str, bool]:
	158	+ if limit <= 0:
	159	+ return "", bool(text)
	160	+ if len(text) <= limit:
	161	+ return text, False
	162	+ return text[:limit].rstrip() + "\n... [truncated]", True
	163	+
	164	+
	165	+def summarize_pr(pr_data: dict[str, Any], files_data: list[dict[str, Any]]) -> dict[str, Any]:
	166	+ additions = sum(int(item.get("additions") or 0) for item in files_data)
	167	+ deletions = sum(int(item.get("deletions") or 0) for item in files_data)
	168	+ return {
	169	+ "title": pr_data.get("title"),
	170	+ "state": pr_data.get("state"),
	171	+ "author": (pr_data.get("user") or {}).get("login"),
	172	+ "created_at": pr_data.get("created_at"),
	173	+ "updated_at": pr_data.get("updated_at"),
	174	+ "draft": bool(pr_data.get("draft")),
	175	+ "mergeable": pr_data.get("mergeable"),
	176	+ "base_branch": ((pr_data.get("base") or {}).get("ref")),
	177	+ "head_branch": ((pr_data.get("head") or {}).get("ref")),
	178	+ "base_repo": (((pr_data.get("base") or {}).get("repo") or {}).get("full_name")),
	179	+ "head_repo": (((pr_data.get("head") or {}).get("repo") or {}).get("full_name")),
	180	+ "labels": [label.get("name") for label in pr_data.get("labels") or [] if label.get("name")],
	181	+ "files_changed": len(files_data),
	182	+ "additions": additions,
	183	+ "deletions": deletions,
	184	+ }
	185	+
	186	+
	187	+def categorize_file(filename: str) -> set[str]:
	188	+ lower = filename.lower()
	189	+ base = os.path.basename(lower)
	190	+ categories: set[str] = set()
	191	+
	192	+ if (
	193	+ lower == "cmakelists.txt"
	194	+ or lower.endswith("/cmakelists.txt")
	195	+ or lower.endswith(".cmake")
	196	+ or lower.startswith("cmake/")
	197	+ or "/cmake/" in lower
	198	+ or lower.endswith(".mk")
	199	+ or base.startswith("build_")
	200	+ or (lower.endswith(".sh") and ("build" in base or "cmake" in base))
	201	+ ):
	202	+ categories.add("build")
	203	+
	204	+ if (
	205	+ lower.startswith("tests/")
	206	+ or "/tests/" in lower
	207	+ or lower.startswith("test/")
	208	+ or "/test/" in lower
	209	+ or base.startswith("test_")
	210	+ or base.endswith("_test.cc")
	211	+ or base.endswith("_test.cpp")
	212	+ or base.endswith("_ut.cc")
	213	+ or base.endswith("_ut.cpp")
	214	+ ):
	215	+ categories.add("tests")
	216	+
	217	+ if lower.endswith((".md", ".rst", ".adoc")):
	218	+ categories.add("docs")
	219	+
	220	+ if lower.endswith((".yaml", ".yml", ".json", ".toml", ".ini", ".cfg", ".conf")):
	221	+ categories.add("config")
	222	+
	223	+ if ".gitlab-ci" in lower or "jenkins" in lower or "workflow" in lower:
	224	+ categories.add("ci")
	225	+
	226	+ if "/third_party/" in lower or lower.startswith("third_party/"):
	227	+ categories.add("third_party")
	228	+
	229	+ if lower.endswith((".c", ".cc", ".cpp", ".cxx", ".h", ".hpp", ".py", ".java", ".go", ".rs")):
	230	+ categories.add("source")
	231	+
	232	+ if not categories:
	233	+ categories.add("other")
	234	+
	235	+ return categories
	236	+
	237	+
	238	+def extract_test_signal(pr_body: str, files_data: list[dict[str, Any]]) -> dict[str, Any]:
	239	+ has_test_files = any("tests" in categorize_file(str(item.get("filename") or "")) for item in files_data)
	240	+ body = (pr_body or "").strip()
	241	+ lower = body.lower()
	242	+
	243	+ if not body:
	244	+ return {"status": "missing", "has_test_files": has_test_files}
	245	+ if re.search(r"(^\|[\s:：])na($\|[\s，。,.\n])", lower) or "不涉及" in body or "无" in body:
	246	+ return {"status": "not_provided", "has_test_files": has_test_files}
	247	+ return {"status": "provided", "has_test_files": has_test_files}
	248	+
	249	+
	250	+def build_review(pr_data: dict[str, Any], files_data: list[dict[str, Any]], summary: dict[str, Any]) -> dict[str, Any]:
	251	+ if not pr_data or not files_data:
	252	+ return {
	253	+ "decision": "insufficient_data",
	254	+ "decision_label": "信息不足",
	255	+ "severity": "unknown",
	256	+ "score": None,
	257	+ "summary": "未获取到完整文件变更，无法生成有效的快速检视结论。",
	258	+ "positives": [],
	259	+ "concerns": ["需要先获取 PR 文件列表和 diff。"],
	260	+ "next_actions": ["重新运行时不要使用 --skip-files，并确认可访问 GitCode API。"],
	261	+ "area_counts": {},
	262	+ "status_counts": {},
	263	+ "top_files": [],
	264	+ "test_signal": {"status": "missing", "has_test_files": False},
	265	+ }
	266	+
	267	+ area_counts: collections.Counter[str] = collections.Counter()
	268	+ status_counts: collections.Counter[str] = collections.Counter()
	269	+ deleted_files = 0
	270	+ rename_only_files = 0
	271	+
	272	+ for item in files_data:
	273	+ filename = str(item.get("filename") or "")
	274	+ status = infer_file_status(item)
	275	+ status_counts[status] += 1
	276	+ if status == "deleted":
	277	+ deleted_files += 1
	278	+ if status == "renamed" and int(item.get("additions") or 0) == 0 and int(item.get("deletions") or 0) == 0:
	279	+ rename_only_files += 1
	280	+ for category in categorize_file(filename):
	281	+ area_counts[category] += 1
	282	+
	283	+ top_files = sorted(
	284	+ (
	285	+ {
	286	+ "filename": str(item.get("filename") or ""),
	287	+ "status": infer_file_status(item),
	288	+ "additions": int(item.get("additions") or 0),
	289	+ "deletions": int(item.get("deletions") or 0),
	290	+ }
	291	+ for item in files_data
	292	+ ),
	293	+ key=lambda item: (item["additions"] + item["deletions"], item["deletions"]),
	294	+ reverse=True,
	295	+ )[:5]
	296	+
	297	+ test_signal = extract_test_signal(str(pr_data.get("body") or ""), files_data)
	298	+ touches_build = area_counts["build"] > 0
	299	+ touches_source = area_counts["source"] > 0
	300	+ docs_only = area_counts["docs"] == len(files_data)
	301	+ non_doc_files = len(files_data) - area_counts["docs"]
	302	+ config_only = (area_counts["docs"] + area_counts["config"]) == len(files_data)
	303	+ total_churn = int(summary["additions"]) + int(summary["deletions"])
	304	+
	305	+ score = 0
	306	+ positives: list[str] = []
	307	+ concerns: list[str] = []
	308	+ next_actions: list[str] = []
	309	+
	310	+ if summary["files_changed"] > 30:
	311	+ score += 2
	312	+ concerns.append(f"改动文件较多，共 {summary['files_changed']} 个文件，人工逐项确认成本高。")
	313	+ elif summary["files_changed"] > 10:
	314	+ score += 1
	315	+ concerns.append(f"改动跨越 {summary['files_changed']} 个文件，已超出轻量改动范围。")
	316	+
	317	+ if total_churn > 1000:
	318	+ score += 3
	319	+ concerns.append(f"代码 churn 很大，新增/删除合计 {total_churn} 行。")
	320	+ elif total_churn > 200:
	321	+ score += 2
	322	+ concerns.append(f"代码 churn 偏大，新增/删除合计 {total_churn} 行。")
	323	+ elif total_churn > 50:
	324	+ score += 1
	325	+
	326	+ if deleted_files > 10:
	327	+ score += 2
	328	+ concerns.append(f"删除文件较多，共 {deleted_files} 个，需确认没有误删构建资产。")
	329	+ elif deleted_files > 0 and summary["deletions"] > 50:
	330	+ score += 1
	331	+ concerns.append(f"存在 {deleted_files} 个删除文件，建议核对删除范围和依赖关系。")
	332	+
	333	+ if touches_build:
	334	+ score += 2
	335	+ concerns.append("触及 CMake/构建脚本/三方库编排，构建回归风险高于普通代码改动。")
	336	+
	337	+ if area_counts["ci"] > 0:
	338	+ score += 2
	339	+ concerns.append("触及 CI/流水线相关文件，需要额外验证流程是否仍然可运行。")
	340	+
	341	+ if touches_source:
	342	+ score += 1
	343	+
	344	+ if test_signal["status"] != "provided" and (touches_build or touches_source):
	345	+ score += 1
	346	+ concerns.append("PR 描述里没有给出有效测试步骤，但改动涉及构建或源码。")
	347	+ next_actions.append("补充可复现的构建/测试步骤，再决定是否 LGTM。")
	348	+
	349	+ if test_signal["has_test_files"]:
	350	+ score -= 1
	351	+ positives.append("PR 同时修改了测试相关文件。")
	352	+
	353	+ if "ci-pipeline-passed" in summary["labels"]:
	354	+ score -= 1
	355	+ positives.append("已有 `ci-pipeline-passed` 标签。")
	356	+
	357	+ if docs_only:
	358	+ score -= 3
	359	+ positives.append("改动仅限文档，风险较低。")
	360	+ elif config_only and not touches_build and not touches_source:
	361	+ score -= 2
	362	+ positives.append("改动主要集中在文档/配置，未触及核心源码。")
	363	+
	364	+ if rename_only_files == len(files_data) and total_churn == 0:
	365	+ score -= 2
	366	+ positives.append("改动主要是纯重命名，没有实际 diff。")
	367	+
	368	+ if summary["mergeable"] is True:
	369	+ positives.append("GitCode 标记为可合并。")
	370	+
	371	+ if score <= 0:
	372	+ decision = "lgtm"
	373	+ decision_label = "可直接 LGTM"
	374	+ severity = "low"
	375	+ short_summary = "从改动规模和类型看属于低风险变更，可以直接给出 LGTM。"
	376	+ next_actions.append("可直接评论字面量 `/lgtm`，如有需要再补一句简短正向反馈。")
	377	+ elif score <= 3:
	378	+ decision = "manual_review"
	379	+ decision_label = "建议人工复查"
	380	+ severity = "medium"
	381	+ short_summary = "当前更像中等风险改动，建议在人工 spot-check 后再决定是否 LGTM。"
	382	+ next_actions.append("优先抽查 churn 最大的 3 个文件，再决定是否评论字面量 `/lgtm`。")
	383	+ else:
	384	+ decision = "do_not_lgtm"
	385	+ decision_label = "不建议直接 LGTM"
	386	+ severity = "high"
	387	+ short_summary = "改动规模或触达范围偏大，当前不适合在未进一步核对前直接 LGTM。"
	388	+ next_actions.append("先核对关键删除文件、CMake 路径调整和实际构建结果，再决定是否放行。")
	389	+
	390	+ if not positives:
	391	+ positives.append("未发现明显的低风险信号。")
	392	+
	393	+ top_areas = [name for name, _count in area_counts.most_common(3)]
	394	+ if top_areas:
	395	+ short_summary += " 主要影响区域：" + "、".join(top_areas) + "。"
	396	+
	397	+ return {
	398	+ "decision": decision,
	399	+ "decision_label": decision_label,
	400	+ "severity": severity,
	401	+ "score": score,
	402	+ "summary": short_summary,
	403	+ "positives": positives,
	404	+ "concerns": concerns[:5],
	405	+ "next_actions": next_actions[:3],
	406	+ "area_counts": dict(area_counts),
	407	+ "status_counts": dict(status_counts),
	408	+ "top_files": top_files,
	409	+ "test_signal": test_signal,
	410	+ }
	411	+
	412	+
	413	+def render_review_markdown(result: dict[str, Any]) -> str:
	414	+ parsed = result["parsed"]
	415	+ lines: list[str] = []
	416	+ lines.append("# CANN PR 快速检视")
	417	+ lines.append("")
	418	+
	419	+ if result["errors"]:
	420	+ lines.append("## 错误")
	421	+ for error in result["errors"]:
	422	+ lines.append(f"- {error}")
	423	+ lines.append("")
	424	+ return "\n".join(lines).rstrip() + "\n"
	425	+
	426	+ review = result.get("review") or {}
	427	+ summary = result.get("summary")
	428	+ lines.append(f"## 结论")
	429	+ lines.append(f"- 判定: {review.get('decision_label', '信息不足')}")
	430	+ lines.append(f"- 严重性: {review.get('severity', 'unknown')}")
	431	+ if review.get("score") is not None:
	432	+ lines.append(f"- 风险分: {review['score']}")
	433	+ lines.append(f"- PR: {parsed['project_path']}#{parsed['pr_number']}")
	434	+ lines.append("")
	435	+ lines.append(review.get("summary") or "未生成摘要。")
	436	+ lines.append("")
	437	+
	438	+ if summary:
	439	+ lines.append("## 概览")
	440	+ lines.append(f"- 标题: {summary['title']}")
	441	+ lines.append(f"- 作者: {summary['author']}")
	442	+ lines.append(f"- 分支: {summary['head_branch']} -> {summary['base_branch']}")
	443	+ lines.append(f"- 文件数: {summary['files_changed']}")
	444	+ lines.append(f"- 增删行: +{summary['additions']} / -{summary['deletions']}")
	445	+ lines.append(f"- 标签: {', '.join(summary['labels']) if summary['labels'] else '(none)'}")
	446	+ lines.append("")
	447	+
	448	+ if review.get("concerns"):
	449	+ lines.append("## 主要关注点")
	450	+ for item in review["concerns"]:
	451	+ lines.append(f"- {item}")
	452	+ lines.append("")
	453	+
	454	+ if review.get("positives"):
	455	+ lines.append("## 正向信号")
	456	+ for item in review["positives"]:
	457	+ lines.append(f"- {item}")
	458	+ lines.append("")
	459	+
	460	+ if review.get("top_files"):
	461	+ lines.append("## 重点文件")
	462	+ for item in review["top_files"][:3]:
	463	+ lines.append(
	464	+ f"- {item['filename']} ({item['status']}, +{item['additions']} / -{item['deletions']})"
	465	+ )
	466	+ lines.append("")
	467	+
	468	+ if review.get("next_actions"):
	469	+ lines.append("## 建议动作")
	470	+ for item in review["next_actions"]:
	471	+ lines.append(f"- {item}")
	472	+ lines.append("")
	473	+
	474	+ if result["notes"]:
	475	+ lines.append("## 备注")
	476	+ for note in result["notes"]:
	477	+ lines.append(f"- {note}")
	478	+ lines.append("")
	479	+
	480	+ actions = result.get("actions") or {}
	481	+ action_lines: list[str] = []
	482	+ if actions.get("comment_posted"):
	483	+ action_lines.append(f"已发布审查评论，comment_id={actions.get('comment_id')}")
	484	+ if actions.get("literal_lgtm_posted"):
	485	+ action_lines.append(
	486	+ f"已按显式请求发布字面量 `/lgtm`，comment_id={actions.get('literal_lgtm_comment_id')}"
	487	+ )
	488	+ if actions.get("lgtm_posted"):
	489	+ action_lines.append(f"已发布字面量 `/lgtm`，comment_id={actions.get('lgtm_comment_id')}")
	490	+ for item in actions.get("skipped") or []:
	491	+ action_lines.append(item)
	492	+ if action_lines:
	493	+ lines.append("## 执行结果")
	494	+ for item in action_lines:
	495	+ lines.append(f"- {item}")
	496	+ lines.append("")
	497	+
	498	+ return "\n".join(lines).rstrip() + "\n"
	499	+
	500	+
	501	+def render_report_markdown(result: dict[str, Any]) -> str:
	502	+ parsed = result["parsed"]
	503	+ review = result.get("review") or {}
	504	+ summary = result.get("summary") or {}
	505	+ now_text = datetime.now().astimezone().strftime("%Y-%m-%d %H:%M:%S %z")
	506	+ severity = str(review.get("severity") or "unknown").lower()
	507	+ severity_label = {
	508	+ "low": "✅ Low",
	509	+ "medium": "⚠️ Medium",
	510	+ "high": "❌ High",
	511	+ "critical": "🔴 Critical",
	512	+ }.get(severity, "❓ Unknown")
	513	+ conclusion = {
	514	+ "lgtm": "✅ 建议合入",
	515	+ "manual_review": "⚠️ 建议人工复查后再决定",
	516	+ "do_not_lgtm": "❌ 当前不建议直接 LGTM",
	517	+ "insufficient_data": "⚠️ 信息不足，无法得出可靠结论",
	518	+ }.get(review.get("decision"), "⚠️ 信息不足")
	519	+ test_signal = (review.get("test_signal") or {}).get("status", "missing")
	520	+ test_label = {
	521	+ "provided": "已提供",
	522	+ "not_provided": "未提供有效测试步骤",
	523	+ "missing": "未提供",
	524	+ }.get(test_signal, "未知")
	525	+ top_files = review.get("top_files") or []
	526	+ top_areas = ", ".join((review.get("area_counts") or {}).keys()) or "(none)"
	527	+ limitations = []
	528	+ files_meta = result.get("files_meta") or {}
	529	+ if files_meta.get("omitted_files", 0) > 0:
	530	+ limitations.append(f"仅展开了部分文件 patch，仍有 {files_meta['omitted_files']} 个文件未在报告中逐条展开。")
	531	+ if result["notes"]:
	532	+ limitations.extend(result["notes"])
	533	+ if not limitations:
	534	+ limitations.append("报告基于 GitCode PR 元数据、文件列表和 patch 摘要生成。")
	535	+
	536	+ lines: list[str] = []
	537	+ lines.append("## 🤖 CANN 代码审查报告")
	538	+ lines.append("")
	539	+ lines.append(f"PR: {parsed['project_path']}#{parsed['pr_number']} - {summary.get('title', '(unknown)')}")
	540	+ lines.append(f"严重性: {severity_label}")
	541	+ lines.append(f"审查时间: {now_text}")
	542	+ lines.append("")
	543	+ lines.append("---")
	544	+ lines.append("")
	545	+ lines.append("### 📊 审查结论")
	546	+ lines.append("")
	547	+ lines.append(f"{conclusion}")
	548	+ lines.append("")
	549	+ lines.append(f"- 严重性: {review.get('severity', 'unknown')}")
	550	+ lines.append(f"- 风险分: {review.get('score', 'n/a')}")
	551	+ lines.append(f"- 文件规模: {summary.get('files_changed', 0)} 个文件，+{summary.get('additions', 0)} / -{summary.get('deletions', 0)}")
	552	+ lines.append(f"- 测试信息: {test_label}")
	553	+ lines.append("")
	554	+ lines.append(review.get("summary") or "未生成摘要。")
	555	+ lines.append("")
	556	+ lines.append("---")
	557	+ lines.append("")
	558	+ lines.append("### 📋 修改概述")
	559	+ lines.append("")
	560	+ lines.append(f"- 作者: {summary.get('author', '(unknown)')}")
	561	+ lines.append(f"- 分支: {summary.get('head_branch', '(unknown)')} -> {summary.get('base_branch', '(unknown)')}")
	562	+ lines.append(f"- 主要影响区域: {top_areas}")
	563	+ lines.append(f"- 标签: {', '.join(summary.get('labels') or []) or '(none)'}")
	564	+ lines.append("")
	565	+ lines.append("---")
	566	+ lines.append("")
	567	+ lines.append("### 🔍 主要关注点")
	568	+ lines.append("")
	569	+ for item in review.get("concerns") or ["未识别到明显关注点。"]:
	570	+ lines.append(f"- {item}")
	571	+ lines.append("")
	572	+ lines.append("---")
	573	+ lines.append("")
	574	+ lines.append("### ✅ 正向信号")
	575	+ lines.append("")
	576	+ for item in review.get("positives") or ["未识别到明显正向信号。"]:
	577	+ lines.append(f"- {item}")
	578	+ lines.append("")
	579	+ if top_files:
	580	+ lines.append("---")
	581	+ lines.append("")
	582	+ lines.append("### 📁 重点文件")
	583	+ lines.append("")
	584	+ for item in top_files:
	585	+ lines.append(
	586	+ f"- `{item['filename']}`: {item['status']}，+{item['additions']} / -{item['deletions']}"
	587	+ )
	588	+ lines.append("")
	589	+ lines.append("---")
	590	+ lines.append("")
	591	+ lines.append("### 💡 建议动作")
	592	+ lines.append("")
	593	+ for item in review.get("next_actions") or ["建议人工补充核查后再处理。"]:
	594	+ lines.append(f"- {item}")
	595	+ lines.append("")
	596	+ lines.append("---")
	597	+ lines.append("")
	598	+ lines.append("### 📝 审查说明")
	599	+ lines.append("")
	600	+ for item in limitations:
	601	+ lines.append(f"- {item}")
	602	+ lines.append("")
	603	+
	604	+ actions = result.get("actions") or {}
	605	+ action_lines: list[str] = []
	606	+ if actions.get("comment_posted"):
	607	+ action_lines.append(f"已发布审查评论，comment_id={actions.get('comment_id')}")
	608	+ if actions.get("literal_lgtm_posted"):
	609	+ action_lines.append(
	610	+ f"已按显式请求发布字面量 `/lgtm`，comment_id={actions.get('literal_lgtm_comment_id')}"
	611	+ )
	612	+ if actions.get("lgtm_posted"):
	613	+ action_lines.append(f"已发布字面量 `/lgtm`，comment_id={actions.get('lgtm_comment_id')}")
	614	+ for item in actions.get("skipped") or []:
	615	+ action_lines.append(item)
	616	+ if action_lines:
	617	+ lines.append("---")
	618	+ lines.append("")
	619	+ lines.append("### 🚀 执行结果")
	620	+ lines.append("")
	621	+ for item in action_lines:
	622	+ lines.append(f"- {item}")
	623	+ lines.append("")
	624	+
	625	+ return "\n".join(lines).rstrip() + "\n"
	626	+
	627	+
	628	+def build_file_context(
	629	+ files_data: list[dict[str, Any]],
	630	+ max_files: int,
	631	+ max_patch_chars: int,
	632	+ max_patch_chars_per_file: int,
	633	+) -> tuple[list[dict[str, Any]], dict[str, Any]]:
	634	+ selected = files_data[: max(0, max_files)]
	635	+ remaining_patch_budget = max(0, max_patch_chars)
	636	+ rendered_files: list[dict[str, Any]] = []
	637	+ omitted_files = max(0, len(files_data) - len(selected))
	638	+
	639	+ for item in selected:
	640	+ patch_text = extract_patch(item)
	641	+ per_file_limit = min(max_patch_chars_per_file, remaining_patch_budget)
	642	+ patch_excerpt, truncated = clip_text(patch_text, per_file_limit)
	643	+ remaining_patch_budget = max(0, remaining_patch_budget - min(len(patch_text), per_file_limit))
	644	+
	645	+ rendered_files.append(
	646	+ {
	647	+ "filename": item.get("filename"),
	648	+ "status": infer_file_status(item),
	649	+ "additions": int(item.get("additions") or 0),
	650	+ "deletions": int(item.get("deletions") or 0),
	651	+ "previous_path": (item.get("patch") or {}).get("old_path") if isinstance(item.get("patch"), dict) else None,
	652	+ "too_large": bool(item.get("too_large")),
	653	+ "patch_included": bool(patch_excerpt),
	654	+ "patch_truncated": truncated,
	655	+ "patch_excerpt": patch_excerpt,
	656	+ }
	657	+ )
	658	+
	659	+ meta = {
	660	+ "selected_files": len(rendered_files),
	661	+ "omitted_files": omitted_files,
	662	+ "total_files": len(files_data),
	663	+ "remaining_patch_budget": remaining_patch_budget,
	664	+ }
	665	+ return rendered_files, meta
	666	+
	667	+
	668	+def render_context_markdown(result: dict[str, Any]) -> str:
	669	+ lines: list[str] = []
	670	+ parsed = result["parsed"]
	671	+ lines.append("# GitCode PR Context")
	672	+ lines.append("")
	673	+ lines.append("## Parsed URL")
	674	+ lines.append(f"- Source URL: {parsed['source_url']}")
	675	+ lines.append(f"- Project: {parsed['project_path']}")
	676	+ lines.append(f"- PR Number: {parsed['pr_number']}")
	677	+ lines.append(f"- Input Kind: {parsed['input_kind']}")
	678	+ lines.append(f"- Canonical Web URL: {parsed['canonical_web_url']}")
	679	+ lines.append(f"- API URL: {parsed['api_url']}")
	680	+ lines.append("")
	681	+ lines.append("## Auth")
	682	+ lines.append(f"- Token Env: {result['auth']['token_env']}")
	683	+ lines.append(f"- Token Present: {'yes' if result['auth']['token_present'] else 'no'}")
	684	+ lines.append(f"- Token Source: {result['auth']['token_source']}")
	685	+ lines.append("")
	686	+
	687	+ if result["notes"]:
	688	+ lines.append("## Notes")
	689	+ for note in result["notes"]:
	690	+ lines.append(f"- {note}")
	691	+ lines.append("")
	692	+
	693	+ if result["errors"]:
	694	+ lines.append("## Errors")
	695	+ for error in result["errors"]:
	696	+ lines.append(f"- {error}")
	697	+ lines.append("")
	698	+
	699	+ if result.get("pr"):
	700	+ pr = result["pr"]
	701	+ summary = result["summary"]
	702	+ lines.append("## Remote Metadata")
	703	+ lines.append(f"- Title: {summary['title']}")
	704	+ lines.append(f"- State: {summary['state']}")
	705	+ lines.append(f"- Author: {summary['author']}")
	706	+ lines.append(f"- Base: {summary['base_repo']} @ {summary['base_branch']}")
	707	+ lines.append(f"- Head: {summary['head_repo']} @ {summary['head_branch']}")
	708	+ lines.append(f"- Mergeable: {summary['mergeable']}")
	709	+ lines.append(f"- Labels: {', '.join(summary['labels']) if summary['labels'] else '(none)'}")
	710	+ lines.append(f"- Files Changed: {summary['files_changed']}")
	711	+ lines.append(f"- Additions / Deletions: {summary['additions']} / {summary['deletions']}")
	712	+ lines.append(f"- Canonical API html_url: {pr.get('html_url')}")
	713	+ lines.append("")
	714	+ if pr.get("body"):
	715	+ lines.append("## PR Body")
	716	+ lines.append("```markdown")
	717	+ lines.append(str(pr["body"]).rstrip())
	718	+ lines.append("```")
	719	+ lines.append("")
	720	+
	721	+ if result.get("files"):
	722	+ meta = result["files_meta"]
	723	+ lines.append("## Changed Files")
	724	+ lines.append(f"- Included: {meta['selected_files']}")
	725	+ lines.append(f"- Omitted: {meta['omitted_files']}")
	726	+ lines.append("")
	727	+ for index, file_item in enumerate(result["files"], start=1):
	728	+ lines.append(f"### {index}. {file_item['filename']}")
	729	+ lines.append(f"- Status: {file_item['status']}")
	730	+ lines.append(f"- Additions / Deletions: {file_item['additions']} / {file_item['deletions']}")
	731	+ if file_item.get("previous_path"):
	732	+ lines.append(f"- Previous Path: {file_item['previous_path']}")
	733	+ if file_item["too_large"]:
	734	+ lines.append("- Patch: marked too_large by GitCode")
	735	+ elif file_item["patch_included"]:
	736	+ lines.append("```diff")
	737	+ lines.append(file_item["patch_excerpt"].rstrip())
	738	+ lines.append("```")
	739	+ else:
	740	+ lines.append("- Patch: omitted")
	741	+ lines.append("")
	742	+
	743	+ return "\n".join(lines).rstrip() + "\n"
	744	+
	745	+
	746	+def build_result(args: argparse.Namespace) -> tuple[dict[str, Any], int]:
	747	+ parsed = parse_pr_url(args.pr_url)
	748	+ token, token_source = resolve_token(args.token_env)
	749	+
	750	+ result: dict[str, Any] = {
	751	+ "parsed": parsed,
	752	+ "auth": {
	753	+ "token_env": args.token_env,
	754	+ "token_present": bool(token),
	755	+ "token_source": token_source,
	756	+ },
	757	+ "summary": None,
	758	+ "review": None,
	759	+ "pr": None,
	760	+ "files": [],
	761	+ "files_meta": None,
	762	+ "actions": {
	763	+ "comment_posted": False,
	764	+ "comment_id": None,
	765	+ "lgtm_posted": False,
	766	+ "lgtm_comment_id": None,
	767	+ "literal_lgtm_posted": False,
	768	+ "literal_lgtm_comment_id": None,
	769	+ "skipped": [],
	770	+ },
	771	+ "notes": [],
	772	+ "errors": [],
	773	+ }
	774	+
	775	+ if not token:
	776	+ result["notes"].append(
	777	+ f"{args.token_env} is not set and no fallback config file was found. URL parsing completed, but PR metadata and file diffs were not fetched."
	778	+ )
	779	+ return result, 0
	780	+ if token_source.startswith("config:"):
	781	+ result["notes"].append(f"Loaded GitCode token from {token_source[7:]}.")
	782	+
	783	+ try:
	784	+ pr_data = fetch_json(parsed["api_url"], token, args.timeout)
	785	+ files_data: list[dict[str, Any]] = []
	786	+ if not args.skip_files:
	787	+ raw_files = fetch_json(parsed["files_api_url"], token, args.timeout)
	788	+ if isinstance(raw_files, list):
	789	+ files_data = raw_files
	790	+ else:
	791	+ raise GitCodeApiError("GitCode files API did not return a list")
	792	+
	793	+ result["pr"] = pr_data
	794	+ result["summary"] = summarize_pr(pr_data, files_data)
	795	+ result["review"] = build_review(pr_data, files_data, result["summary"])
	796	+ result["parsed"]["canonical_web_url"] = pr_data.get("html_url") or result["parsed"]["canonical_web_url"]
	797	+ files_context, files_meta = build_file_context(
	798	+ files_data,
	799	+ max_files=args.max_files,
	800	+ max_patch_chars=args.max_patch_chars,
	801	+ max_patch_chars_per_file=args.max_patch_chars_per_file,
	802	+ )
	803	+ result["files"] = files_context
	804	+ result["files_meta"] = files_meta
	805	+ return result, 0
	806	+ except GitCodeApiError as exc:
	807	+ result["errors"].append(str(exc))
	808	+ return result, 1
	809	+
	810	+
	811	+def execute_actions(args: argparse.Namespace, result: dict[str, Any]) -> int:
	812	+ if not args.post_comment and not args.auto_lgtm and not args.lgtm_only:
	813	+ return 0
	814	+
	815	+ if result["errors"]:
	816	+ result["actions"]["skipped"].append("存在错误，未执行评论或 LGTM 动作。")
	817	+ return 1
	818	+
	819	+ if not result.get("pr") or not result.get("summary"):
	820	+ result["actions"]["skipped"].append("未获取到完整 PR 信息，未执行评论或 LGTM 动作。")
	821	+ return 1
	822	+
	823	+ token, _token_source = resolve_token(args.token_env)
	824	+ if not token:
	825	+ result["actions"]["skipped"].append("缺少 GitCode token，未执行评论或 LGTM 动作。")
	826	+ return 1
	827	+
	828	+ parsed = result["parsed"]
	829	+ review = result.get("review") or {}
	830	+
	831	+ try:
	832	+ if args.post_comment:
	833	+ if args.comment_format == "review":
	834	+ comment_body = render_review_markdown(result)
	835	+ else:
	836	+ comment_body = render_report_markdown(result)
	837	+ comment_response = post_comment(parsed["comments_api_url"], token, args.timeout, comment_body)
	838	+ result["actions"]["comment_posted"] = True
	839	+ result["actions"]["comment_id"] = comment_response.get("id")
	840	+
	841	+ if args.lgtm_only:
	842	+ lgtm_response = post_comment(parsed["comments_api_url"], token, args.timeout, "/lgtm")
	843	+ result["actions"]["literal_lgtm_posted"] = True
	844	+ result["actions"]["literal_lgtm_comment_id"] = lgtm_response.get("id")
	845	+
	846	+ if args.auto_lgtm:
	847	+ if review.get("decision") == "lgtm":
	848	+ lgtm_response = post_comment(parsed["comments_api_url"], token, args.timeout, "/lgtm")
	849	+ result["actions"]["lgtm_posted"] = True
	850	+ result["actions"]["lgtm_comment_id"] = lgtm_response.get("id")
	851	+ else:
	852	+ result["actions"]["skipped"].append("当前判定不是“可直接 LGTM”，已跳过自动 LGTM 命令。")
	853	+ return 0
	854	+ except GitCodeApiError as exc:
	855	+ result["errors"].append(str(exc))
	856	+ return 1
	857	+
	858	+
	859	+def parse_args() -> argparse.Namespace:
	860	+ parser = argparse.ArgumentParser(
	861	+ description="Normalize GitCode PR URLs and fetch PR context for Codex/Claude skills."
	862	+ )
	863	+ parser.add_argument("pr_url", help="GitCode PR URL")
	864	+ parser.add_argument("--format", choices=("review", "report", "json", "context-markdown", "markdown"), default="review")
	865	+ parser.add_argument("--token-env", default="GITCODE_TOKEN")
	866	+ parser.add_argument("--timeout", type=float, default=20.0)
	867	+ parser.add_argument("--max-files", type=int, default=40)
	868	+ parser.add_argument("--max-patch-chars", type=int, default=16000)
	869	+ parser.add_argument("--max-patch-chars-per-file", type=int, default=4000)
	870	+ parser.add_argument("--skip-files", action="store_true")
	871	+ parser.add_argument("--post-comment", action="store_true")
	872	+ parser.add_argument("--comment-format", choices=("review", "report"), default="report")
	873	+ parser.add_argument("--auto-lgtm", action="store_true")
	874	+ parser.add_argument("--lgtm-only", action="store_true")
	875	+ parser.add_argument("--output", help="Optional output file path")
	876	+ return parser.parse_args()
	877	+
	878	+
	879	+def main() -> int:
	880	+ args = parse_args()
	881	+ try:
	882	+ result, exit_code = build_result(args)
	883	+ except ValueError as exc:
	884	+ error_result = {
	885	+ "errors": [str(exc)],
	886	+ "notes": [],
	887	+ }
	888	+ output = (
	889	+ json.dumps(error_result, ensure_ascii=False, indent=2)
	890	+ if args.format == "json"
	891	+ else "# CANN PR 快速检视\n\n## 错误\n- " + str(exc) + "\n"
	892	+ )
	893	+ if args.output:
	894	+ with open(args.output, "w", encoding="utf-8") as handle:
	895	+ handle.write(output)
	896	+ else:
	897	+ sys.stdout.write(output)
	898	+ return 2
	899	+
	900	+ action_exit_code = execute_actions(args, result)
	901	+
	902	+ if args.format == "json":
	903	+ output = json.dumps(result, ensure_ascii=False, indent=2)
	904	+ elif args.format in {"context-markdown", "markdown"}:
	905	+ output = render_context_markdown(result)
	906	+ elif args.format == "report":
	907	+ output = render_report_markdown(result)
	908	+ else:
	909	+ output = render_review_markdown(result)
	910	+
	911	+ if args.output:
	912	+ with open(args.output, "w", encoding="utf-8") as handle:
	913	+ handle.write(output)
	914	+ else:
	915	+ sys.stdout.write(output)
	916	+ return max(exit_code, action_exit_code)
	917	+
	918	+
	919	+if __name__ == "__main__":
	920	+ sys.exit(main())