The Agent Skills Directory

[COMMAND_EXECUTION]: The skill implements and utilizes tools to execute shell commands, which are necessary for the workshop's development tasks like running the project (uv run main.py) or performing code analysis (rg).
The exec_workspace tool found in references/reference_agent_core.py and references/starter_main_wg21.py uses subprocess.run with shell=True to execute commands.
This implementation includes safety measures such as a blocked list (e.g., rm -rf, del /f) and workspace root enforcement via resolve_workspace_path.
[EXTERNAL_DOWNLOADS]: The skill relies on the uv package manager to handle Python dependencies. This involves downloading software packages from the Python Package Index (PyPI) or similar official registries, which is standard behavior for the project's use-case.
[DATA_EXFILTRATION]: No patterns of unauthorized data exfiltration were detected. The skill reads local project files and environment configurations to assist the student, but it does not contain logic to send this information to external or unknown network destinations.
[PROMPT_INJECTION]: The instructions in SKILL.md use directive language to maintain the workshop's state machine and routing logic. No malicious prompt injection patterns, such as attempts to bypass AI safety filters or exfiltrate system instructions, were identified.

peas-workshop-advanced-coach