The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requests permission to use Bash for installing required Python dependencies via pip and uv, and for interacting with the Claude CLI. This functionality is essential for its stated purpose of scaffolding and initializing new agent projects.\n- [EXTERNAL_DOWNLOADS]: The skill provides templates and instructions that involve downloading official packages from the Python Package Index (PyPI) and utilizing npx to execute well-known Model Context Protocol (MCP) servers, such as the official Notion integration. These references target established, trusted ecosystems.\n- [PROMPT_INJECTION]: As an architectural scaffolding tool, the skill ingests user project descriptions to generate system prompts and orchestration logic. This creates a surface for indirect prompt injection if the user-provided project specifications originate from untrusted sources; however, the skill explicitly provides templates for implementing safety hooks and guardrails to help users secure their generated systems.

agent-architect