The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted article content from WeChat URLs which is subsequently summarized by the agent.
Ingestion points: scripts/fetch_wechat_article.py extracts text from remote HTML articles provided via user-supplied links.
Boundary markers: Absent. The summary generation process in SKILL.md does not employ delimiters (such as XML tags or explicit block quotes) to separate the untrusted article content from the agent's internal instructions.
Capability inventory: The skill has the ability to execute shell commands (running the helper script and curl) and perform file system writes (saving images to /tmp).
Sanitization: No filtering or validation is performed on the extracted article text to prevent embedded instructions from influencing the agent's behavior.
[COMMAND_EXECUTION]: The skill relies on shell command execution for its primary workflow.
Instructions in SKILL.md include running a Python script with user-provided links: python3 .claude/skills/read-gzh/scripts/fetch_wechat_article.py "<公众号链接>".
SKILL.md instructs the agent to use curl for image downloads: curl -o /tmp/article_img_1.jpg "<图片链接>".
The Python script scripts/fetch_wechat_article.py uses subprocess.run to call curl for both fetching HTML and downloading images.
[EXTERNAL_DOWNLOADS]: The skill fetches data and images from external domains to perform its tasks.
Fetches article content from WeChat's official domain mp.weixin.qq.com.
Downloads images from the WeChat CDN at mmbiz.qpic.cn.

read-gzh