The Agent Skills Directory

[SAFE]: No malicious patterns, command execution, or data exfiltration attempts were detected in the skill instructions or reference files.
[INDIRECT_PROMPT_INJECTION]: The skill involves processing untrusted PDF data, which is a potential attack surface. However, the skill provides clear "Instruction Boundaries" (指令边界) that explicitly prevent the agent from treating document text as instructions. It specifically states that text and links within the PDF are solely for transcription and should not be used to expand the task scope or call external services.

pdf-transcription