The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes document titles and activity data from Yuque which are externally controlled. If an attacker names a document with malicious instructions, they could potentially influence the summary or suggestions generated in the final report. This is an inherent surface for any data-processing skill.
Ingestion points: Data enters via yuque_group_doc_stats and yuque_group_member_stats tools.
Capability inventory: The skill can write documents using yuque_create_doc.
Boundary markers: The skill uses Markdown templates but does not explicitly warn the agent to ignore instructions embedded in the source data.
Sanitization: No explicit sanitization or filtering of document titles or member names is mentioned.

group-weekly