Official Account Assistant

Security checks across malware telemetry and agentic risk

Overview

This skill can help with drafts, but it asks for WeChat publishing/API authority and mass-send automation without clear scoping, and it promotes hiding AI authorship.

Treat this skill as a draft-writing aid only unless it adds clear account-permission declarations and mandatory final approval before any publish or mass send. Do not authorize WeChat API access or run missing helper scripts without reviewing them, and avoid using it to hide AI authorship in ways that could mislead readers or violate platform rules.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Risk analysis

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.

#
ASI02: Tool Misuse and Exploitation
High
What this means

The agent could publish or mass-send content to followers in a way that affects the user's public account and reputation.

Why it was flagged

This directs the agent toward public-account publishing and mass messaging, which are high-impact actions, but the artifacts do not clearly bound final approval, audience scope, reversibility, or a safe draft-only default.

Skill content
支持公众号后台自动发布;定时推送功能;自动排版优化;一键群发
Recommendation

Use this only for draft preparation unless the skill adds explicit preview, per-post approval, audience confirmation, logging, and cancellation/rollback guidance.

#
ASI03: Identity and Privilege Abuse
High
What this means

Granting broad account/API access could let the agent operate a publishing account beyond the user's expectations.

Why it was flagged

The skill asks for WeChat Official Account login and API authorization, but the registry metadata declares no primary credential, required environment variable, config path, or scope boundary for that account access.

Skill content
Step 1:登录公众号 - 扫码登录微信公众平台 - 授权 API 访问权限
Recommendation

Do not authorize the account until the skill clearly declares required credentials, exact API scopes, storage/handling rules, and requires user confirmation before any publish or mass send.

#
ASI09: Human-Agent Trust Exploitation
Medium
What this means

Readers may be led to believe AI-generated or AI-assisted content was fully human-written, creating reputational and compliance risk.

Why it was flagged

The skill explicitly promotes concealing AI authorship, and it pairs that with public-account publishing workflows, which can mislead readers or violate platform/content policies.

Skill content
彻底去除 AI 痕迹,根本看不出是 AI 写的
Recommendation

Use humanization for readability, not deception; avoid fabricated personal experiences and disclose AI assistance when required by platform rules or audience expectations.

#
ASI04: Agentic Supply Chain Vulnerabilities
Low
What this means

If a user later obtains or runs similarly named helper scripts, their behavior may differ from the reviewed artifact set.

Why it was flagged

These referenced helper scripts are not included in the supplied file manifest, so the claimed auto-publish and image-generation implementations are not reviewable here.

Skill content
`scripts/remove_ai_flavor.py` - AI 降味脚本; `scripts/auto_publish.py` - 自动发布脚本; `scripts/generate_images.py` - 智能配图脚本
Recommendation

Do not run additional scripts for publishing or image generation unless their source is included, reviewed, and pinned to a trusted version.

#
ASI07: Insecure Inter-Agent Communication
Low
What this means

Unpublished article ideas or sensitive draft details could be shared with third-party image services.

Why it was flagged

If implemented, this feature may send article content or prompts to external image-generation providers; the artifacts do not define provider identity, retention, or data handling, although this is aligned with the stated image-generation purpose.

Skill content
根据文章内容自动生成配图; 支持 DALL-E 3、Midjourney 等 AI 绘图
Recommendation

Avoid sending confidential drafts to image providers unless you understand and accept their terms, retention practices, and credential handling.

#
ASI06: Memory and Context Poisoning
Low
What this means

Draft titles and content may remain on disk after use and could be reused or viewed later.

Why it was flagged

The code stores created article records and draft content in a persistent local JSON file under the skill directory, which is expected for article management but should be visible to users.

Skill content
ARTICLES_FILE = DATA_DIR / "articles.json" ... data["articles"].append(article) ... json.dump(data, f, ensure_ascii=False, indent=2)
Recommendation

Do not store confidential drafts unless local retention is acceptable, and delete articles.json when you no longer need the saved drafts.