The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill recommends using npx commands to download and execute bootstrap scripts directly from a GitHub repository for environment setup and configuration generation.
Evidence: npx --yes github:sandraschi/advanced-memory-mcp/scripts/bootstrap/windows in modules/ai-ide-development.md and similar commands for macOS and Linux in modules/distribution-and-installation.md.
Note: These are vendor resources originating from the skill author's repository.
[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for building tools that ingest untrusted data from external sources and perform sensitive operations, creating an indirect injection surface.
Ingestion points: comprehensive_research tool in modules/core-guidance.md fetches data from web, academic, and news sources; collect_data in modules/core-guidance.md.
Boundary markers: The guide emphasizes structured JSON responses but does not explicitly require delimiters (like triple backticks or XML tags) for external content interpolation in the agent prompt.
Capability inventory: The guide includes implementations for network requests (aiohttp), file system operations (aiofiles), and mentions shell command execution in modules/architecture-and-standards.md.
Sanitization: The skill strongly recommends Pydantic for input validation and shlex.quote for shell command sanitization.
[EXTERNAL_DOWNLOADS]: The documentation references and recommends downloading resources from several external domains and repositories.
Fetches documentation and release notes from github.com/anthropics/anthropic-sdk and github.com/modelcontextprotocol/specification.
References MCP marketplaces and registries including glama.ai, lobehub.com, skillsmp.com, and mcp.cool.
[COMMAND_EXECUTION]: The skill provides instructions for executing shell commands for project initialization, linting, and testing.
Evidence: Commands such as pip install, mcpb init, uv sync, and uv run ruff check are scattered throughout the guide.

mcp-server-architect