varg-occasion-video

Pass

Audited by Gen Agent Trust Hub on Nov 1, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied text and images that are subsequently used in generation prompts for various AI models.
  • Ingestion points: Occasion details (names, dates, places) and user photos are ingested as described in SKILL.md.
  • Boundary markers: Instructions recommend using quotes and specific spelling commands (e.g., "spelled exactly like that") to help the model distinguish input text from instructions.
  • Capability inventory: Capabilities include network requests via curl, file system manipulation (Read, Write, Edit), and media processing via ffmpeg in scripts/assemble.sh.
  • Sanitization: No explicit sanitization or filtering is applied to the input text before inclusion in generation prompts, relying on manual agent and user review as a final check.
  • [COMMAND_EXECUTION]: Orchestrates video assembly using a local Bash script scripts/assemble.sh that wraps ffmpeg and ffprobe for media manipulation.
  • [EXTERNAL_DOWNLOADS]: Communicates with official vendor endpoints at api.varg.ai and mcp.varg.ai to invoke content generation models.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 1, 2026, 12:01 PM