The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as community posts, comments, and WeCom chat logs through various search and detail retrieval commands.
Ingestion points: External data is ingested via retrieval commands defined in references/search_posts.md, references/get_post_detail.md, and references/community-chat-analysis.md.
Boundary markers: The skill utilizes structured report templates (e.g., in references/community-daily-report.md) but does not specify explicit prompt delimiters for raw user content.
Capability inventory: The skill uses the ae-cli tool to perform read operations and write operations, including reporting data to external endpoints via ae-cli community data report.
Sanitization: The skill provides clear instructions for filtering content, such as removing emoticons or system messages (e.g., in references/community-analyzing-theme-comment.md), and guidelines for masking sensitive identifiers like user IDs and phone numbers (in references/community-chat-analysis.md).

ae-community