ai-summarizing
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION] (LOW): Indirect Prompt Injection surface detected in examples.md.
- Ingestion points: Untrusted data enters the agent context via input fields: 'transcript' in ProcessMeeting, 'thread' in SummarizeThread, and 'section' in SummarizeSection (all in examples.md).
- Boundary markers: Absent. The signatures do not define delimiters (e.g., XML tags or triple quotes) to separate untrusted data from the rest of the prompt.
- Capability inventory: The skill performs LLM-based summarization, structured data extraction, and iterative chunk processing across all examples.
- Sanitization: Absent. There is no logic to escape, filter, or validate the content of the external text inputs before they are interpolated into the model's context.
Audit Metadata