character-design-sheet
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill configuration uses the allowed-tools property to restrict the agent's shell access. Only commands starting with 'belt' are permitted, preventing arbitrary command execution.
- [COMMAND_EXECUTION]: Bash commands are used to invoke the belt CLI for image generation tasks. These commands include authentication ('belt login') and running inference apps ('belt app run'), which are consistent with the skill's primary purpose.
- [EXTERNAL_DOWNLOADS]: The skill references an external documentation file for the CLI installation process located at a GitHub repository under the inference-sh organization. This is a common practice for providing setup instructions.
- [INDIRECT_PROMPT_INJECTION]: The skill structures tasks where user-provided descriptions are interpolated into CLI command arguments. The security risk is mitigated by the restricted command set and the intended use case of image generation.
Audit Metadata