← Back to Skills Marketplace
263
Downloads
0
Stars
1
Active Installs
1
Versions
Install in OpenClaw
/install privacy-protector
Description
Runs PII anonymization, local de-anonymization, and deterministic local detector checks for text and supported files. Use for redact/restore flows, file-firs...
Usage Guidance
This skill appears to do what it says: local `lite` mode runs offline, and other modes send text to a remote anonymization API. If you must keep data strictly local, use `--level lite` only. Before using `dynamic`/`strict`/`crossborder`, confirm you trust the default backend (https://safety-cf.modeio.ai) or explicitly set ANONYMIZE_API_URL to a trusted endpoint. Be aware that the tool writes map files (which include original values) to ~/.modeio/redact/maps by default — secure or relocate that directory if it will contain sensitive data. If you need higher assurance, review the remote API provider code (modeio_redact/providers/remote_api_provider.py) and test on non-sensitive examples first.
Capability Analysis
Type: OpenClaw Skill
Name: privacy-protector
Version: 0.1.0
The privacy-protector skill bundle is a PII anonymization and de-anonymization utility that supports various file formats including .docx and .pdf. It features a local 'lite' mode using regex and a remote mode that communicates with a stated backend API (safety-cf.modeio.ai) for advanced processing. The implementation follows security best practices, such as enforcing strict filesystem permissions (0o600) on local mapping data stored in the user's home directory (~/.modeio/redact/maps). No evidence of malicious intent, unauthorized data exfiltration, or prompt injection was found in the code or instructions.
Capability Assessment
Purpose & Capability
Name/description match the implementation: Python scripts and a reusable package implement anonymize, deanonymize, and local detection. Required binary is only python3 and there are no unrelated credentials or unexpected binaries requested.
Instruction Scope
SKILL.md and CLI scripts are scoped to anonymize/deanonymize/detect workflows and instruct running commands inside the repo. The README and SKILL.md explicitly state that only `lite` is fully local; other levels call a backend API. The instructions do not ask the agent to read system files outside the repo, but runtime will read/write local map files (default ~/.modeio/redact/maps).
Install Mechanism
No install spec is provided (instruction-only install); the repository includes Python code but nothing in the manifest downloads or executes remote installers. Runtime depends on optional Python packages (requests, python-docx, PyMuPDF) which the user must install themselves.
Credentials
The skill does not require secrets or credentials. However, non-`lite` anonymization levels will POST content to an external API by default (URL hardcoded as https://safety-cf.modeio.ai/api/cf/anonymize) and an environment variable (ANONYMIZE_API_URL) can override that endpoint. Maps containing original PII are written to a default local directory (~/ .modeio/redact/maps) unless overridden. Both behaviors are justified by the skill's purpose but increase data-exfiltration risk if you don't trust the backend or if ANONYMIZE_API_URL is set to an attacker-controlled host.
Persistence & Privilege
The skill is not always-enabled and does not request elevated platform privileges. It persists local map files and sidecar files in a user-writable directory, which is expected behavior for restore/redact workflows.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install privacy-protector - After installation, invoke the skill by name or use
/privacy-protector - Provide required inputs per the skill's parameter spec and get structured output
Version History
v0.1.0
Initial release of privacy-protector.
- Adds local PII anonymization, de-anonymization, and local detector diagnostics for text and supported files.
- Supports redact/restore file workflows with allowlist, blocklist, and threshold configuration.
- Fully offline processing for the "lite" anonymization level; backend API support for other levels.
- Includes helpers for working with text, various file types, and saved placeholder maps.
- Provides documentation for usage patterns, supported commands, and runtime considerations.
Metadata
Frequently Asked Questions
What is Privacy Protector?
Runs PII anonymization, local de-anonymization, and deterministic local detector checks for text and supported files. Use for redact/restore flows, file-firs... It is an AI Agent Skill for Claude Code / OpenClaw, with 263 downloads so far.
How do I install Privacy Protector?
Run "/install privacy-protector" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Privacy Protector free?
Yes, Privacy Protector is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Privacy Protector support?
Privacy Protector is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Privacy Protector?
It is built and maintained by modeioai (@modeioai); the current version is v0.1.0.
More Skills