wechat-mp-article-publisher
Pass
Audited by Gen Agent Trust Hub on Aug 8, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or unauthorized access attempts were detected. The skill's behavior is consistent with its stated purpose.
- [COMMAND_EXECUTION]: The skill executes browser automation commands (e.g.,
browser navigate,browser act) to interact with the WeChat Official Account backend. All operations are targeted at the legitimate domainmp.weixin.qq.com. - [PROMPT_INJECTION]: Analysis of the attack surface for Indirect Prompt Injection:
- Ingestion points: The skill ingests untrusted user data for the article title, body, and AI illustration prompts (found in
SKILL.md). - Boundary markers: None are present; user content is directly interpolated into browser typing commands.
- Capability inventory: The skill possesses browser interaction capabilities, including navigation, snapshots, and clicking/typing actions.
- Sanitization: There is no explicit sanitization or escaping of user-provided article content.
- Contextual Mitigation: The risk is significantly mitigated by mandatory human-in-the-loop checkpoints. The platform requires manual QR code scans by the administrator for both the initial login and the final publishing confirmation, preventing the agent from performing high-stakes actions without direct user authorization.
Audit Metadata