The Agent Skills Directory

[DATA_EXFILTRATION]: The skill processes sensitive authentication cookies (ZHIHU_COOKIES). The implementation in scripts/blog.py includes a defensive cookie_header function that validates the target domain before attaching cookies, ensuring they are only sent to the intended Zhihu API endpoints.
[COMMAND_EXECUTION]: The skill executes a local Python script (scripts/blog.py) that uses only the Python standard library to communicate with the Zhihu API. The script implements a safety mechanism where write operations (publishing) require an explicit --confirm flag as the final argument, preventing accidental or unauthorized posts.
[DATA_EXPOSURE]: Instructions in SKILL.md explicitly caution the agent against printing or echoing the ZHIHU_COOKIES environment variable, which contains full account access credentials.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the Zhihu platform (article titles and excerpts) which could contain malicious instructions.
Ingestion points: Data is fetched from Zhihu APIs via the articles and article commands in scripts/blog.py.
Boundary markers: Not present; the data is returned as raw JSON objects.
Capability inventory: The skill has access to the Bash tool and can perform authenticated network writes to the user's Zhihu account.
Sanitization: No sanitization is performed on the article content before it is processed by the agent.

cn-blog