NYC

wechat-article-publisher

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFECOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [CREDENTIALS_UNSAFE] (HIGH): The skill instructions direct the agent to execute cat .env | grep WECHAT_API_KEY. This action reads sensitive credentials from a file and places them directly into the agent's conversation context, increasing the risk of credential leakage through logs or subsequent prompt injections.
  • [COMMAND_EXECUTION] (MEDIUM): The skill relies on the execution of local scripts (wechat_api.py and parse_markdown.py) located in ~/.claude/skills/wechat-article-publisher/scripts/. The source code for these scripts was not provided, making their behavior unverifiable.
  • [DATA_EXFILTRATION] (LOW): The skill is designed to send article content and authentication tokens to https://wx.limyai.com. While this is the primary function of the skill, the destination is a non-whitelisted third-party domain, posing a potential risk if the service is compromised.
  • [INDIRECT_PROMPT_INJECTION] (LOW): The skill processes untrusted Markdown and HTML files provided by users.
  • Ingestion points: Path arguments passed to wechat_api.py (e.g., /path/to/article.md).
  • Boundary markers: None identified; user-provided file content is directly parsed.
  • Capability inventory: Subprocess execution of Python scripts and network POST requests to an external API.
  • Sanitization: No evidence of input sanitization or validation of the file content before processing.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 04:56 PM