lovstudio-review-doc — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited lovstudio-review-doc (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
AI-powered document and contract review. Outputs annotated docx with comments and/or tracked changes.
审阅文档或合同，AI 进行审查分析，将审查意见以批注（comment）或修订模式 （track changes）写回文档，输出带批注的 docx。
用户提供文档要求审阅、审查、批注、批阅
合同/协议审查：识别风险条款、权利义务失衡、模糊表述、缺失条款
通用文档审阅：语法、逻辑、格式、内容完整性
Extract paragraph text and indices / 用脚本提取段落文本和索引：
python3 lovstudio-review-doc/scripts/annotate_docx.py extract --input <path.docx>
Outputs JSON array, each item has index (paragraph number) and text. Includes text inside tracked changes (<w:ins>) — e.g. counterparty revisions visible in Track Changes mode. 输出 JSON 数组，每项含 index（段落序号）和 text（含修订追踪内容）。
Important: If the document has tracked changes (e.g. Schedule A added via Track Changes), these paragraphs are now included in extraction. Always review the full extracted output before generating annotations. 如果文档包含修订追踪内容（如对方以 Track Changes 添加的附件），这些段落 现已包含在提取结果中。
IMPORTANT: Use `AskUserQuestion` BEFORE generating annotations.
Review mode / 审阅模式确认
━━━ Review type / 审阅类型 ━━━ a) Contract review — risk clauses, rights, vague wording, missing clauses 合同/协议审查 — 风险条款、权利义务、模糊表述、缺失条款 b) General document review — grammar, logic, formatting, completeness 通用文档审阅 — 语法、逻辑、格式、内容完整性
━━━ Annotation mode / 批注方式 ━━━ a) Comments only (default) — add comments alongside text, no changes 批注模式（默认） — 在原文旁加 comment，不改原文 b) Track changes — directly modify text, recipient can accept/reject 修订模式 — 用 track changes 直接改原文，对方可逐条接受/拒绝 c) Comments + track changes — comments for analysis, revisions for suggestions 批注 + 修订 — 批注写分析意见，修订写建议改法
━━━ Author name / 批注者署名 ━━━ Default "手工川", customizable
Review the extracted text. For contract review, focus on:
根据提取的文本进行审查。对于合同审查，重点关注：
生成 annotations JSON，格式：
{ "comments": [ { "paragraph": 18, "text": "【风险】Sourced Deal 认定...", "author": "手工川" } ], "revisions": [ { "paragraph": 12, "old": "terminates automatically", "new": "terminates automatically, provided that...", "author": "手工川" } ] }
Annotation text format / 批注文本格式规范：
【风险/Risk】、【建议/Suggestion】、【缺失/Missing】、【模糊/Vague】、【注意/Note】将 JSON 写入临时文件，调用脚本：
python3 lovstudio-review-doc/scripts/annotate_docx.py annotate \ --input <原文.docx> \ --annotations <annotations.json> \ --output <输出路径.docx>
输出文件名规范：手工川-{原文件名}-审阅-{YYYY-MM-DD}-v0.1.docx
放在原文件同目录下。
python3 annotate_docx.py extract --input <path.docx> python3 annotate_docx.py annotate --input <path.docx> --annotations <json> --output <path.docx>
| Subcommand | Argument | Description |
|---|---|---|
extract | --input | 输入 docx 路径 |
annotate | --input | 输入 docx 路径 |
annotate | --annotations | JSON 批注文件路径 |
annotate | --output | 输出 docx 路径 |
| Field | Required | Description |
|---|---|---|
paragraph | Yes | 0-based 段落索引 |
text | Yes | 批注内容 |
author | No | 批注者署名（默认 "Reviewer"） |
start | No | 字符偏移起始位置（精确高亮） |
end | No | 字符偏移结束位置 |
| Field | Required | Description |
|---|---|---|
paragraph | Yes | 0-based 段落索引 |
old | Yes | 要替换的原文 |
new | Yes | 修改后的文本 |
author | No | 修订者署名 |
old text in revisions must match a single<w:r> (run) in the paragraph. If the target text is split across multiple runs (common with smart quotes, spell-check, or mixed formatting), the revision will be skipped. Workaround: match a substring within one run, or use a comment instead. 修订的 old 文本必须完整匹配段落中的单个 run。如果目标文本跨 run，修订会 被跳过。可改用批注替代。
and starts new IDs after it. No manual ID management needed. 脚本自动检测已有批注的最大 ID，新批注从其后开始编号。
pip install python-docx --break-system-packages
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.