The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill acts on external transcript data provided by the user, which constitutes an indirect prompt injection surface. While malicious instructions could be embedded in transcripts, the skill lacks any capabilities (such as network access, file system manipulation, or tool execution) to be exploited. 1. Ingestion points: Transcript notes or text provided for the ANALYZE mode. 2. Boundary markers: None specified; the instructions do not suggest using delimiters to separate input from agent instructions. 3. Capability inventory: No tools or shell commands are requested or used. 4. Sanitization: No methods for validating or filtering external transcript content are included.

interview