The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs users to install a command-line interface (CLI) tool using a piped shell command: curl -fsSL https://cli.inference.sh | sh. This is a high-risk pattern that allows for arbitrary code execution from a remote source directly into the user's shell environment.\n- [COMMAND_EXECUTION]: Windows installation procedures for development tools involve bypassing PowerShell's execution policy: powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex". This facilitates the execution of unverified scripts by overriding system security configurations.\n- [EXTERNAL_DOWNLOADS]: The skill facilitates the download of various installation scripts and configurations from external sources:\n
Fetches the uv tool from astral.sh.\n
Downloads the fnm installer from Vercel Labs' infrastructure.\n
Retrieves the nvm installer from its official repository on GitHub.\n
Fetches configuration logic from cli.inference.sh.\n- [PROMPT_INJECTION]: The skill describes building applications that process external data through AppInput schemas, creating a surface for indirect prompt injection. The analysis identifies the following components:\n
Ingestion points: Untrusted data enters the agent context through the input_data object in the run() methods (e.g., in SKILL.md and logic reference files).\n
Boundary markers: The provided templates do not use delimiters or instructions to prevent the agent from obeying instructions embedded in the user input.\n
Capability inventory: The apps described have extensive capabilities, including file system access (writeFileSync, open), logging, and the ability to make network requests to external APIs.\n
Sanitization: There is no evidence of sanitization or validation of the input content before it is processed by the app logic.

building-inferencesh-apps