The Agent Skills Directory

[SAFE]: The skill serves as a development guide and reference repository for the EdgeOne Makers platform. It does not contain executable code itself, but rather provides copy-pasteable templates and architectural guidance.
[COMMAND_EXECUTION]: The skill mentions CLI commands (e.g., edgeone makers dev, edgeone makers deploy) and platform tools (e.g., context.sandbox.runCode). These are documented as standard platform operations for local development and agent functionality within the EdgeOne ecosystem.
[EXTERNAL_DOWNLOADS]: The documentation references standard package registries (NPM, PyPI) for framework dependencies like @anthropic-ai/claude-agent-sdk, langgraph, and crewai. These are well-known services and essential for the stated purpose of the skill.
[PROMPT_INJECTION]: The skill includes defensive instructions for AI agents, such as Red Line #11 (Error handling) and #13 (Sandboxed execution), designed to improve the reliability and safety of the agents built using these templates.
[DATA_EXFILTRATION]: The skill enforces an 'Iron Rule' for environment variables, mandating the use of context.env instead of process.env or os.environ to ensure proper isolation and secret management within the serverless runtime.

makers-agents