product-changelog
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the user to install external tools including the
belt-sh/cliandinference-sh/skillspackages vianpx. It also references installation documentation hosted on a public GitHub repository atraw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md. - [COMMAND_EXECUTION]: The skill provides numerous examples of shell commands using the
beltCLI to run applications such asfalai/flux-dev-lorafor image generation,infsh/agent-browserfor automated screenshots, andbytedance/seededit-3-0-i2ifor image editing. - [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: The skill demonstrates using the
infsh/agent-browsertool, which accepts external URLs (e.g.,https://your-app.com/new-feature) to capture UI screenshots. This allows the agent to process content from potentially untrusted web pages. - Boundary markers: Absent; there are no instructions provided to the agent to delimit or ignore instructions that might be embedded within the content of the target URLs.
- Capability inventory: The skill possesses command execution capabilities through the
beltCLI tool and network interaction through browser automation. - Sanitization: No sanitization or validation mechanisms for the data retrieved from external URLs are described.
Audit Metadata