The Agent Skills Directory

[DATA_EXFILTRATION]: The skill manages authentication via the WEIBO_COOKIES environment variable. The Python implementation includes a domain-matching utility to ensure cookies are only transmitted to weibo.com hosts and utilizes unredirected headers to prevent session tokens from leaking to third-party domains in the event of an API redirect.
[PROMPT_INJECTION]: The skill processes external data from the Weibo AJAX API, creating an indirect prompt injection surface.
Ingestion points: User profile details and recent statuses are fetched via scripts/weibo.py from official Weibo endpoints.
Boundary markers: The script outputs all retrieved data as structured JSON via json.dumps, providing a clear structural boundary between data and instructions for the agent.
Capability inventory: The skill uses Bash for command execution and urllib for authenticated network requests to the target service.
Sanitization: Content is escaped through JSON serialization and status text is truncated for display purposes.
[COMMAND_EXECUTION]: The skill utilizes Bash to run a local Python script. It enforces a manual safety checkpoint for the post command, requiring a specific --confirm flag as the final argument to prevent unauthorized or accidental public social media posts.

weibo