ai-marketing-kit-orchestrator
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions to download a repository from an external source (github.com/crealwork/ai-marketing-kit) and copy its contents into the agent's local skills directory.
- [INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface by processing untrusted data.
- Ingestion points: User input collected during the onboarding interview process is stored in local files (DESIGN.md and BRAND-VOICE.md).
- Boundary markers: The implementation does not demonstrate the use of delimiters to prevent the agent from interpreting user-provided brand data as new instructions.
- Capability inventory: The suite includes capabilities for file writing, making external API calls via the requests library, and executing system commands via ffmpeg and the higgsfield CLI.
- Sanitization: There is no evidence of sanitization or strict schema validation for the user-provided content before it is stored and referenced as brand configuration.
- [COMMAND_EXECUTION]: The orchestration logic relies on invoking external binary tools such as ffmpeg for video processing, faster-whisper for transcription, and the higgsfield CLI for image generation.
Audit Metadata