XHS Writing Coach
ReviewAudited by ClawScan on May 10, 2026.
Overview
This instruction-only writing coach has no executable code, but its reference notes include unbounded persistent memory instructions and a stale rule to remove AI-content labels, so users should review it before use.
Before installing, remove or revise the persistent-memory update instruction, delete the user-specific 'Hervé' preference, and replace the 'no #AI生成内容' rule with a current-user/platform compliance check. Also ask before using Tavily or any external search on private draft material.
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.
The agent could carry old or personal preferences into future drafts, or store new observations in the skill notes without asking.
The skill directs persistent memory updates and already contains a named user's preference, showing that user-specific or stale context can be retained and reused without stated consent, retention limits, or review boundaries.
This file is **the long-term “know-how” memory** ... Update it whenever we learn something new ... Default tone for Hervé
Require explicit user approval before updating any reference file, define exactly what may be stored, avoid personal data, and provide a way to review or delete stored notes.
A user may trust the skill to handle AI-label compliance but receive output that omits a required AI-generated-content label, risking platform or compliance problems.
This hard-coded removal instruction conflicts with the skill's advertised AI-content compliance reminder and appears to reflect a prior user's preference rather than the current user's or platform's requirements.
Tags 5–8, **no #AI生成内容** (user requested removal)
Remove the hard-coded 'no #AI生成内容' rule and make AI-label decisions depend on the current user's instruction and the applicable platform requirement.
If a topic or draft contains private details, those details could be included in an external search query if the agent uses Tavily.
The notes suggest using an external search provider for grounding. This is purpose-aligned for trend research, but the metadata does not declare the provider or describe what user-supplied topics may be sent.
## 5) Hot topic grounding workflow (Tavily) We use Tavily (not Brave). Keep links/citations when possible.
Ask the user before using external search, avoid sending private details in queries, and document the provider dependency in the skill metadata or usage notes.