The Agent Skills Directory

[REMOTE_CODE_EXECUTION] (CRITICAL): The skill instructions include the command curl -fsSL https://cli.inference.sh | sh. This is a high-risk pattern that executes a remote script directly in the local shell environment without verification of the content or the source.
[EXTERNAL_DOWNLOADS] (HIGH): The skill downloads binaries and additional logic from inference.sh and dist.inference.sh domains. These are not included in the trusted provider list, making the skill a potential vector for supply chain attacks.
[COMMAND_EXECUTION] (MEDIUM): The skill requests permission to use the Bash tool with infsh commands. While this is the intended purpose, it provides a large attack surface if the agent is manipulated into running malicious CLI flags or arguments.
[PROMPT_INJECTION] (LOW): The skill's primary function is to interpolate user-provided prompts into CLI command strings (e.g., infsh app run --input '...'). This creates a surface for indirect prompt injection or command injection if inputs are not properly sanitized. (Evidence: Ingestion point at the --input flag in SKILL.md; Boundary markers are limited to single quotes; Capability inventory includes system shell access; Sanitization logic is not provided).

prompt-engineering