meeting-insights-analyzer

Pass

Audited by Gen Agent Trust Hub on Jul 12, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill instructions and metadata contain no signs of obfuscation, malicious command execution, or unauthorized data exfiltration. The functionality is purely analytical and consistent with the stated purpose.
  • [PROMPT_INJECTION]: The skill is subject to an indirect prompt injection surface as it processes external meeting transcripts. Analysis of the attack surface:
  • Ingestion points: The skill scans and reads meeting transcripts in .txt, .md, .vtt, .srt, and .docx formats from user-specified folders.
  • Boundary markers: Absent; the instructions do not explicitly provide delimiters to separate transcript content from agent instructions.
  • Capability inventory: The skill has no capabilities for network access, file writing, or subprocess execution, meaning any potential injection could not result in data exfiltration or system compromise.
  • Sanitization: No validation or sanitization of transcript content is performed. Given the lack of dangerous capabilities, this surface is considered a functional characteristic rather than a security risk.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 12, 2026, 11:39 PM
Security Audit — agent-trust-hub — meeting-insights-analyzer