nmt-analyze-interviews
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process large volumes of untrusted user-provided content, such as interview transcripts, sales calls, and chat logs.
- Ingestion points: Processes external files in formats including .md, .txt, .docx, and .csv through user-provided paths.
- Boundary markers: The skill instructions emphasize the use of a rigid extraction schema and 'honest confidence' levels to distinguish between raw data and synthesized conclusions.
- Capability inventory: Performs local file reads of methodology references and writes a final report to the filesystem; it also uses platform subagents for distributed processing.
- Sanitization: Uses a 'fan-out' architecture where subagents distill raw transcripts into compact structured summaries before synthesis by the orchestrator, reducing the risk of processing raw malicious instructions.
Audit Metadata