product-changelog
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill recommends installing the
beltCLI and related skills from theinference-shorganization. These are well-known resources within the platform's ecosystem. - [COMMAND_EXECUTION]: The skill makes extensive use of the
beltCLI tool to execute remote AI models for image generation (falai/flux-dev-lora), image editing (bytedance/seededit-3-0-i2i), and browser automation (infsh/agent-browser). All commands are standard usage for the platform. - [INDIRECT_PROMPT_INJECTION]: The skill acts as a template for processing project updates into changelogs, presenting a theoretical attack surface if the input source (like git logs or raw commit messages) contains malicious instructions.
- Ingestion points: Project source code, git logs, and release notes (implied context).
- Boundary markers: None identified in the provided templates.
- Capability inventory: Execution of
beltcommands allowed viaBash(belt *)tool, including image generation, stitching, and browser automation. - Sanitization: No specific sanitization or filtering logic is described in the prompt templates.
Audit Metadata