highlight-scanner
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through processed video transcripts.
- Ingestion points: The script
scripts/find_highlights.pyreads external transcript data (SRT/VTT formats) and incorporates the text content into prompts sent to external LLMs (OpenAI/Gemini). - Boundary markers: The prompt template in the
find_highlights_with_aifunction lacks clear delimiters or 'ignore embedded instructions' warnings when interpolating thetranscript_linesvariable. - Capability inventory: The skill has the capability to perform network operations (calling external AI APIs) and provides the results back to the agent environment.
- Sanitization: There is no evidence of sanitization, filtering, or escaping of the transcript content before it is embedded into the prompt, allowing potential instructions inside a transcript to influence the behavior of the external model.
Audit Metadata