AI Customer Service

PassAudited by ClawScan on May 13, 2026.

Overview

This appears to be a benign customer-service policy skill, but it relies on a local wiki and customer chat data that should be kept accurate and protected.

Use this skill if you want a customer-service agent grounded in a local support wiki. Before production use, make sure the configured wiki path is limited to approved support content, wiki edits are controlled, and the WeChat/WeCom adapter protects customer identifiers, messages, and attachments. The provided SKILL.md content is truncated in the review artifact, so inspect the full skill text before relying on it in a live support channel.

Findings (2)

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

Customer replies may repeat whatever is in the local support wiki, including outdated or incorrect policy information.

Why it was flagged

The skill intentionally relies on persistent local knowledge-base and log files to produce customer-facing answers. This is purpose-aligned, but those files can influence future responses if stale, inaccurate, or edited by an untrusted party.

Skill content
Use the local llm-wiki as the source of truth... Read `SCHEMA.md`... Read `index.md`... Read the recent end of `log.md`... search all markdown files
Recommendation

Keep the support wiki curated, restrict write access, review log updates, and scope SUPPORT_WIKI_PATH/WIKI_PATH to the intended support knowledge base only.

What this means

Customer messages, identifiers, and attachments may be processed by the agent runtime and supporting adapter.

Why it was flagged

The skill is designed to operate behind a messaging-channel adapter and handle customer identifiers, messages, and attachments. This is expected for customer support, but the adapter boundary and customer data handling need appropriate validation and privacy controls.

Skill content
customer WeChat → company WeCom/Enterprise WeChat → channel adapter → OpenClaw/Hermes runtime → this skill → local wiki... Capture `channel`, `external_user_id`, `external_message_id`, `conversation_id`, text, attachments, locale
Recommendation

Use a trusted channel adapter, validate message origin and idempotency, minimize stored customer data, and scan or restrict attachments before passing them to the agent.