infsh-cli
Pass
Audited by Gen Agent Trust Hub on Mar 12, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill facilitates CLI installation using
curl -fsSL https://cli.inference.sh | sh, which pipes a remote shell script from the vendor's domain directly to the shell. - [EXTERNAL_DOWNLOADS]: The skill downloads CLI binaries, manifests, and updates from the vendor's distribution infrastructure at
https://dist.inference.sh. - [COMMAND_EXECUTION]: The skill is authorized to execute
infshcommands via theBashtool to perform AI tasks and manage cloud resources. - [DATA_EXFILTRATION]: The CLI tool includes a documented feature to automatically upload local files (such as images, audio, or video) to the cloud when paths are provided in input parameters. While this is a primary feature, it represents a surface for potential data exposure if sensitive paths are used. Ingestion points:
infsh app runcommands accepting local paths inSKILL.mdandreferences/running-apps.md. Boundary markers: Absent. Capability inventory:Bash(infsh *)tool with file-reading and network-upload capabilities. Sanitization: Absent. - [PROMPT_INJECTION]: The skill processes user-supplied JSON inputs that are passed to remote AI models, creating a surface for indirect prompt injection. Ingestion points:
infsh app runinput flags. Boundary markers: Absent. Capability inventory:Bash(infsh *)tool with network and file system access. Sanitization: Absent.
Audit Metadata