The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill downloads Python scripts (main.py) and unit tests from a GitHub repository (github.com/fancyboi999/ai-engineering-from-scratch-zh) and executes them to validate lab exercises.
[COMMAND_EXECUTION]: The agent is instructed to use python3 and the unittest module to run scenario runners, validators, and tests located in the lesson directories.
[EXTERNAL_DOWNLOADS]: The skill fetches configuration files (program.json), certification tracks, lesson documentation, and quizzes from a remote repository owned by the vendor.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from remote JSON and Markdown files, creating an attack surface where malicious content in the repository could influence agent behavior.
Ingestion points: Files including program.json, tracks/*.json, docs/zh.md, and quiz.json are read from a remote GitHub repository at runtime.
Boundary markers: The instructions do not specify the use of delimiters or specific warnings to ignore instructions embedded within the downloaded lesson materials.
Capability inventory: The skill has the ability to execute shell commands (python3) and write to the local file system to maintain student status.
Sanitization: There is no mention of sanitizing or validating the content of the remote Markdown or JSON files before they are interpreted by the agent.
[CREDENTIALS_UNSAFE]: The skill documentation references the use of an ANTHROPIC_API_KEY for optional online testing but explicitly mandates that the key must never be printed, persisted, or stored in source code.

claude-certification