Install
openclaw skills install @zhaoxinghua09-cell/doc-compliance-assistant文档合规处理助手 — 清洗+合规扫描一体:编码探测/BOM/换行/空白归一化后,一次扫描 PII(邮箱/手机号/身份证)、密钥(GitHub/AWS/OpenAI/私钥)、URL 内嵌口令,输出脱敏命中报告。当你要发布/外发/归档文档前做合规自检时使用;触发词:doc-compliance、文档合规、发前扫描、PII 扫描、密钥泄漏检查。
openclaw skills install @zhaoxinghua09-cell/doc-compliance-assistantdoc-compliance-assistant v1.0.0 · LGD-Powered security/productivity family · zero-dependency
You finish a draft, a config dump, a log export — and then you wonder: did I accidentally paste a phone number, a bearer token, or an internal IP into something I'm about to send out? Anthropics' official docx/pdf skills are pure format converters; they never look at content. This skill fuses normalization and compliance into one pass so you don't ship a document and then find the leak.
--json) to wire into CI or a pre-commit hook.python scripts/doc_compliance.py --file path/to/draft.md
python scripts/doc_compliance.py --file path/to/draft.md --json
python scripts/doc_compliance.py --file path/to/draft.md --clean --out normalized.txt
python scripts/doc_compliance.py --list-rules
con***com, ghp***bcd) — never printed in plaintext.doc_compliance.py per Usage. Default is scan-only; add --clean --out to also emit a normalized copy.findings_secret > 0, treat as a block until the source file is scrubbed.$ python scripts/doc_compliance.py --file examples/sample-input.md
file: examples/sample-input.md | encoding: utf-8 | chars: 306
findings: 5 (secret=2 pii=2 info=1)
- [PII] email line=3 match=con***com
- [PII] cn_mobile line=3 match=138***000
- [secret] aws_access_key line=5 match=AKI***PLE
- [secret] github_token line=6 match=ghp***bcd
- [info] ipv4 line=9 match=192***.10
This is a local, offline scanner. It does not call any API, does not upload content, and does not auto-redact. It is a stop-gate, not a DLP pipeline. For regulated pipelines pair it with a real DLP; this skill exists so the cheap 80% case (obvious secrets / PII) never leaves the editor unwatched.
Security aid only. Regex scanners have false negatives; treat a clean run as "no obvious patterns", not "this document is safe".
doc-compliance-assistant v1.0.0 · LGD-Powered 安全合规家族 · 零依赖
草稿、配置导出、日志文件写完了,转头就犯嘀咕:我是不是把手机号、Bearer Token、内网 IP 不小心贴进了马上要发出去的东西里?anthropics 官方 docx/pdf 件只做格式转换,从不看内容。本技能把归一化和合规扫描压成一遍,避免"发出去之后才发现泄漏"。
--json)接 CI 或 pre-commit。python scripts/doc_compliance.py --file path/to/draft.md
python scripts/doc_compliance.py --file path/to/draft.md --json
python scripts/doc_compliance.py --file path/to/draft.md --clean --out normalized.txt
python scripts/doc_compliance.py --list-rules
con***com、ghp***bcd),绝不打印明文。doc_compliance.py。默认只扫描;加 --clean --out 同时输出归一化副本。findings_secret > 0,视为阻断,直到源文件被清干净。$ python scripts/doc_compliance.py --file examples/sample-input.md
file: examples/sample-input.md | encoding: utf-8 | chars: 306
findings: 5 (secret=2 pii=2 info=1)
- [PII] email line=3 match=con***com
- [PII] cn_mobile line=3 match=138***000
- [secret] aws_access_key line=5 match=AKI***PLE
- [secret] github_token line=6 match=ghp***bcd
- [info] ipv4 line=9 match=192***.10
本地、离线扫描。不调任何 API、不上传内容、不自动脱敏。它是一道闸门,不是 DLP 流水线。受监管流水线请叠真 DLP;本技能存在的意义是把"明显密钥/PII"这便宜的 80% 挡在编辑器里。
本工具为安全辅助。正则扫描有误报漏报;rc=0 表示"未命中明显模式",不等于"文档绝对安全"。