infsh-cli

Fail

Audited by Gen Agent Trust Hub on Apr 22, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill provides an installation command curl -fsSL https://cli.inference.sh | sh in SKILL.md and multiple reference files. This executes a remote script from the vendor's domain directly in the shell.
  • [EXTERNAL_DOWNLOADS]: The installation and manual setup processes download binaries, manifests, and checksums from the vendor's distribution server at dist.inference.sh.
  • [DATA_EXFILTRATION]: The CLI automatically identifies local file paths provided in application inputs and uploads them to the vendor's cloud infrastructure (cloud.inference.sh). This creates a surface for indirect prompt injection or data exposure.
  • Ingestion points: Application input JSON provided to the infsh app run command as described in SKILL.md and references/running-apps.md.
  • Boundary markers: No boundary markers or 'ignore' instructions are specified for the application inputs.
  • Capability inventory: The Bash(infsh *) tool has the capability to read local files and perform network uploads as part of its core functionality.
  • Sanitization: No sanitization or path validation is documented to prevent the agent from accessing sensitive files if instructed by a user or malicious input.
  • [COMMAND_EXECUTION]: The skill's primary function involves executing the infsh binary with various arguments to manage and run AI applications.
Recommendations
  • HIGH: Downloads and executes remote code from: https://cli.inference.sh - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Apr 22, 2026, 04:25 PM
Security Audit — agent-trust-hub — infsh-cli