← 返回 Skills 市场
293
总下载
0
收藏
0
当前安装
5
版本数
在 OpenClaw 中安装
/install privacy-guard
功能描述
Protects sensitive files (contracts, transaction data, personal/customer privacy) by performing deep content inspection for text and binary formats (PDF, DOC...
安全使用建议
This skill appears internally coherent for locally scanning files. Before installing: (1) review and, if desired, customize the keyword lists to reduce false positives (the code includes a broader 'Company Confidential' category not noted in the SKILL.md); (2) pin dependency versions in requirements.txt and install in a sandbox or virtualenv; (3) test the scanner on non-sensitive sample files to confirm PASS/BLOCK behavior; (4) remember the SKILL.md policy is advisory — ensure the agent or your integration actually enforces the 'must scan' rule rather than relying on user discipline. If you need stricter guarantees, wrap file-send operations with a hard enforcement layer that calls this scanner and blocks sends when the script returns BLOCK.
功能分析
Type: OpenClaw Skill
Name: privacy-guard
Version: 0.3.0
The 'privacy-guard' skill is a legitimate utility designed to prevent accidental data leakage by scanning files for sensitive keywords (e.g., contracts, PII, financial data) before an agent processes them. The implementation in 'scan_file.py' uses standard libraries (pdfplumber, python-docx, openpyxl) to inspect local files and returns a 'BLOCK' or 'PASS' signal based on content, with no evidence of data exfiltration, obfuscation, or malicious intent.
能力评估
Purpose & Capability
Name/description match the delivered artifacts: a Python-based local scanner and SKILL.md that requires python3 and lists the same parsing libraries. No unrelated binaries, cloud credentials, or external endpoints are requested.
Instruction Scope
SKILL.md mandates always scanning before any read/send and defines BLOCK/PASS semantics; the provided scan_file.py implements those rules. Note: the enforcement is a runtime policy the agent must follow — the skill cannot technically prevent a user/agent from ignoring the rule. Also: the code defines an extra 'Company Confidential' category not listed in the SKILL.md category summary, which is a minor documentation/code mismatch that could affect outcomes.
Install Mechanism
No remote download or installer; installation is via pip using requirements.txt (or an install_deps.sh wrapper). This is standard for Python packages; no obscure URLs or extracted archives are present.
Credentials
The skill declares no environment variables, no credentials, and no config paths. The scanner operates entirely on local files and local Python libraries, so requested privileges are minimal and proportional.
Persistence & Privilege
always is false and the skill does not request persistent/system-wide configuration or modify other skills. Autonomous model invocation is allowed by default but is not combined with broad credentials or other red flags.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install privacy-guard - 安装完成后，直接呼叫该 Skill 的名称或使用
/privacy-guard触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v0.3.0
- Added a Chinese summary to enhance accessibility for Chinese-speaking users.
- Strengthened enforcement language in the core rules, making scanning and blocking steps explicitly mandatory and non-negotiable.
- Emphasized strict blocking policy and unconditionally stopping tasks if sensitive content is detected.
- Clarified that bypassing the restriction is strictly forbidden under any circumstances.
- Specified that validation via the scanning script is required before any data transmission.
v0.2.0
- Added requirements.txt to specify Python dependencies for installation.
- Added scripts/install_deps.sh to simplify environment setup.
- Updated documentation to include installation instructions using these new files.
v0.1.2
- Adds deep content inspection for both text and binary formats, including PDF, DOCX, and XLSX files.
- Expands supported file types to include .txt, .csv, .md, .json, .xml, .log, .py, .js, .html, .pdf, .docx, and .xlsx.
- Lists new external dependencies: pdfplumber, python-docx, and openpyxl.
- Clarifies that sensitive information scanning covers file content in addition to filenames.
v0.1.1
privacy-guard 0.1.1
- Expanded sensitive keyword lists for transaction, personal, and customer privacy detection.
- Clarified that scanning covers both filenames and file contents.
- Updated scanning instructions to use a relative path from the skill's base directory.
- Changed default activation: skill is no longer always-on by default.
- Renamed skill from sense-privacy-guard to privacy-guard.
v0.1.0
Initial release of privacy-guard (sense-privacy-guard):
- Scans local files for sensitive information before any read, search, upload, or send operation.
- Blocks actions when contracts, transaction data, personal, or customer privacy keywords are detected.
- Requires all local files to be validated using a Python scanning script before upload or transmission.
- Clearly informs the user if sensitive content is found and halts further processing to ensure privacy.
- Enforces strict, non-bypassable privacy policy for sensitive data handling.
元数据
常见问题
sense-privacy-guard 是什么？
Protects sensitive files (contracts, transaction data, personal/customer privacy) by performing deep content inspection for text and binary formats (PDF, DOC... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 293 次。
如何安装 sense-privacy-guard？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install privacy-guard」即可一键安装，无需额外配置。
sense-privacy-guard 是免费的吗？
是的，sense-privacy-guard 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
sense-privacy-guard 支持哪些平台？
sense-privacy-guard 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 sense-privacy-guard？
由 yacki（@yacki）开发并维护，当前版本 v0.3.0。
推荐 Skills