Presidio Pii Scrubber for sensitive info

Local PII protection for OpenClaw agents. Scrubs customer data (names, phones, emails, addresses, credit cards, vessel names) before it reaches any AI model....

This skill appears to do what it says: run local Presidio containers, anonymize customer data before sending it to models, then restore locally. Before installing: 1) Ensure you run the Presidio containers from a trusted source (the skill references mcr.microsoft.com/presidio images). 2) Treat the mapping directory (~/.openclaw/presidio/mappings by default) as highly sensitive — restrict access, exclude it from backups if appropriate, and monitor deletion success. 3) Test failure modes: confirm that mapping files are deleted after restore and that blocked behavior triggers when containers are down. 4) If you operate on shared machines or backups, consider changing PRESIDIO_MAPPING_DIR to a secure location and audit it. 5) Review network rules so containers cannot exfiltrate data (the scripts call localhost only but container config matters). If you need, run a short security review or penetration test focused on mapping file handling and container network isolation before deploying in production.

Type: OpenClaw Skill Name: presidio-pii-skill Version: 0.1.1 The presidio-pii skill implements local PII protection by routing sensitive data through Microsoft Presidio Docker containers on localhost. It features scripts for health monitoring (presidio-health.sh), reversible anonymization (presidio-scrub.py), and restoration (presidio-restore.py) using local mapping files with restricted permissions (chmod 600). The logic is transparent, lacks external network calls, and includes a 'fail-closed' policy in SKILL.md to prevent data leakage if the protection service is offline, aligning perfectly with its stated security purpose.