The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install third-party skills using the command npx skills add <owner/repo@skill> -g -y. The use of the -y flag explicitly bypasses user confirmation during the installation of external code from arbitrary GitHub repositories, while the -g flag performs a global installation on the host system, creating a significant supply chain risk.
[COMMAND_EXECUTION]: The skill includes and instructs the agent to execute a local Python script (scripts/detect_stack.py) to scan project directories for dependency information. It also provides instructions for running various CLI tools such as agent-browser, uvx, and npx.
[REMOTE_CODE_EXECUTION]: The instructions for installing MCP servers (e.g., npx -y @modelcontextprotocol/server-* or uvx mcp-server-*) involve downloading and executing code directly from public registries at runtime without integrity verification of the specific packages.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface. The scripts/detect_stack.py script reads and parses content from untrusted project files (such as package.json, pyproject.toml, and requirements.txt). Malicious instructions embedded within these files could be ingested into the agent's context when it reviews the script's output, potentially influencing downstream behavior without proper sanitization.

light-tool-selection