The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user and agent to clone a repository from an unverified external source: https://github.com/LING71671/Open-ClaudeCode.git. This repository is not maintained by a trusted organization or a well-known service provider.
[REMOTE_CODE_EXECUTION]: Following the download, the skill provides instructions to execute the downloaded code using node package/cli.js. This allows the execution of arbitrary scripts from an untrusted external repository.
[REMOTE_CODE_EXECUTION]: The project contains pre-compiled binaries in package/vendor/ripgrep/ and package/vendor/audio-capture/. The skill guides the agent to use a CLI that relies on these opaque binaries, which originate from an unverified source and could contain malicious payloads.
[COMMAND_EXECUTION]: The skill explicitly documents and encourages the use of the --dangerously-skip-permissions flag. This flag is designed to bypass built-in safety prompts and authorization checks for file system and system operations, increasing the risk of autonomous malicious activity.
[COMMAND_EXECUTION]: The provided integration examples use child_process.exec to invoke the CLI programmatically. If the prompt variable passed to this function is not strictly sanitized, it creates a vector for command injection attacks.

open-claudecode-research