The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches educational content (JSON manifests and Markdown documentation) from a repository belonging to the skill author (fancyboi999/ai-engineering-from-scratch-zh). This is considered normal vendor functionality for a tutorial-based skill.
[COMMAND_EXECUTION]: The skill executes Python 3 scripts for labs (e.g., python3 phases/13-tools-and-protocols/06-mcp-fundamentals/code/main.py) and checks for a TypeScript runner. These actions are scoped to the local repository directory and are intended for educational purposes.
[DATA_EXFILTRATION]: While the skill accesses local project files and reads environment information (like python3 --version), it explicitly instructs the agent to redact sensitive information such as tokens, secrets, cookies, and authorization headers from its logs (MCP-LEARNING.md). It does not send data to unauthorized external endpoints; network operations are limited to fetching course materials from the author's GitHub repository.
[PRIVILEGE_ESCALATION]: The skill includes a "Public deployment gate" which forces the agent to stop and request explicit user confirmation before any network binding or deployment activities, effectively implementing a least-privilege guardrail.

learn-mcp