find-quotes
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external search results and transcript data from third-party YouTube and podcast sources.
- Ingestion points: Untrusted content enters the agent's context through the
arcmira.searchandarcmira.transcriptfunctions as defined inSKILL.md. - Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are utilized to prevent the agent from obeying commands hidden within the transcript text.
- Capability inventory: The skill environment includes the
arcmira_execute_readtool, allowing the agent to execute program logic based on fetched data. - Sanitization: The transcript text is mapped directly to output fields without visible sanitization or filtering for adversarial instructions.
- [DYNAMIC_EXECUTION]: The skill utilizes the
arcmira_execute_readtool to run JavaScript programs that orchestrate entity resolution, search filtering, and transcript processing.
Audit Metadata