The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to perform workspace management tasks, such as creating the architecture-studio-101 directory and copying sandbox files. These operations are restricted to the learner's local practice folder.
[EXTERNAL_DOWNLOADS]: The skill instructions specify retrieving original sandbox resources and templates from the author's repository at github.com/AlpacaLabsLLC during the setup phase. These downloads are limited to the specific project files needed for the course.
[INDIRECT_PROMPT_INJECTION]: The course exercises involve processing unstructured and "deliberately messy" data, such as voice memo transcripts and unverified CSV files. The skill addresses this potential surface by teaching defensive habits, including the "plan first, build second" approach and requiring the agent to trace claims back to specific source lines.
[DYNAMIC_EXECUTION]: In Module 5, the learner is guided to manually create a new skill file (SKILL.md) in a subfolder of the practice directory. This is used as a pedagogical exercise to demonstrate how the AI platform triggers and follows procedures, rather than for malicious persistent execution.

learn