The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The SKILL.md file contains explicit instructions for the agent to verify the API key by running cat .env | grep WECHAT_API_KEY. This command prints the sensitive credential into the agent's context and user-visible output, which is a security risk for secret management.
[DATA_EXFILTRATION]: The script wechat_api.py is configured to send user-supplied content (articles) along with the WECHAT_API_KEY to the external domain https://wx.limyai.com. This domain is a third-party service that is not recognized as a well-known or trusted service in the analyzer's configuration.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests and processes untrusted external document content.
Ingestion points: The skill reads local Markdown and HTML files through arguments provided to wechat_api.py and parse_markdown.py.
Boundary markers: Absent. There are no delimiters or specific instructions provided to the agent or the API to treat the file content as purely passive data.
Capability inventory: The skill combines file-reading capabilities with network-outbound capabilities to an external API.
Sanitization: The parse_markdown.py script performs structural conversion to HTML but does not implement security sanitization to filter out potential malicious instructions or payloads within the processed text.

wechat-article-publisher