The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions facilitate the ingestion of untrusted data from external QQ Channel forum threads and comments, creating a potential surface for indirect prompt injection.
Ingestion points: External data is retrieved from the QQ Open Platform via GET /channels/{channel_id}/threads and GET /channels/{channel_id}/threads/{thread_id} endpoints defined in SKILL.md and references/api_references.md.
Boundary markers: The instructions do not define specific delimiters or "ignore instructions" directives to ensure the agent disregards commands that might be embedded within the retrieved forum content.
Capability inventory: The skill utilizes the qqbot_platform_api tool which has the capability to perform operations including POST, PUT, PATCH, and DELETE on channels, forum posts, and schedules.
Sanitization: There are no instructions provided to sanitize, filter, or validate the content of posts or comments before the agent processes or displays them.

qqbot-channel