Openclaw Pii Anonymizer Latest

ReviewAudited by ClawScan on May 10, 2026.

Overview

This PII anonymizer mostly matches its privacy purpose, but its script sends private text to Ollama using unsafe JSON construction that could break or bypass anonymization.

Use this only with a trusted local Ollama endpoint, and do not rely on it as a complete privacy barrier until the script safely JSON-escapes input. Check that jq, curl, Ollama, and the phi3:mini model are installed, since the registry metadata does not declare all setup requirements.

Findings (3)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

Text containing quotes, backslashes, or crafted JSON-like content could cause the anonymization request to fail or behave differently, reducing confidence that PII is actually scrubbed.

Why it was flagged

The script inserts user-controlled text directly into a JSON string passed to curl without JSON encoding or escaping.

Skill content
{\"role\": \"user\", \"content\": \"$input\"}
Recommendation

Build the request body with a JSON encoder such as jq --arg, send input via stdin or a safely escaped field, and test inputs containing quotes, newlines, and backslashes.

What this means

Private memory or workspace snippets may be sent to the configured Ollama endpoint before being reused in other model or tool workflows.

Why it was flagged

The skill is intended to process memory and tool-call content, which may contain sensitive or persistent context.

Skill content
Use for memory_search, tool calls, HEARTBEAT.md sanitization
Recommendation

Keep OLLAMA_URL pointed only at a local or trusted Ollama service, limit the amount of memory sent for sanitization, and review anonymized output before sending it externally.

What this means

The skill may not work as expected unless jq, curl, Ollama, and the phi3:mini model are installed and configured manually.

Why it was flagged

SKILL.md declares runtime dependencies that the registry metadata does not declare, and the package has no install spec.

Skill content
requires: { bins: [jq, curl], env: [OLLAMA_URL] }
Recommendation

Verify dependencies manually before use, and prefer a package version whose registry metadata accurately declares required binaries, environment variables, and setup steps.