zoho

Pass

Audited by Gen Agent Trust Hub on Apr 17, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads meeting recordings in MP4 format from Zoho's official public and private file servers (e.g., files-accl.zohopublic.com) as documented in the meeting recording pipeline.
  • [COMMAND_EXECUTION]: The skill executes local shell commands via the zoho CLI and standup-summarizer.sh script, including curl for API interactions, jq for data parsing, and ffmpeg for extracting audio from video files.
  • [DATA_EXFILTRATION]: Extracted audio data from meeting recordings is transmitted to Google's Gemini API (generativelanguage.googleapis.com) for transcription. This is a legitimate data flow required for the meeting summarization feature.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted data from meeting transcriptions.
  • Ingestion points: Meeting audio recordings and transcripts processed by scripts/standup-summarizer.sh and documented in SKILL.md.
  • Boundary markers: None identified in the transcription processing logic.
  • Capability inventory: The agent has the ability to read, create, and modify records in Zoho CRM and Projects using the bin/zoho CLI tool.
  • Sanitization: No explicit sanitization or filtering of transcription content is performed before summarization.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 17, 2026, 11:13 PM