claymation-ad

Pass

Audited by Gen Agent Trust Hub on Aug 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes local tools such as ffmpeg and Whisper for media analysis and assembly. It also executes internal scripts including scripts/setup.sh, scripts/check-novoads-env.sh, and scripts/music_mix.py to manage the environment and finalize production. These are identified as vendor-owned resources from the author 'novoads'.\n- [PROMPT_INJECTION]: The skill presents an indirect prompt injection vulnerability because it ingests untrusted data from user-provided video files to recreate ads.\n
  • Ingestion points: The workflow for analyzing found videos in SKILL.md involves reading external files to extract dialogue and beat structures.\n
  • Boundary markers: The instructions do not specify delimiters or provide warnings to the agent to ignore instructions that might be embedded in the processed video content.\n
  • Capability inventory: The skill has local execution privileges for ffmpeg, file-writing capabilities for assembly, and network access to the Novoads REST API.\n
  • Sanitization: No sanitization or validation of the content extracted from external videos is performed before it is interpolated into subsequent prompts.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 14, 2026, 12:34 AM
Security Audit — agent-trust-hub — claymation-ad