The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill manages a course workflow that ingests both local user-modifiable files and remote external content, creating a surface for potential instruction injection.
Ingestion points: The skill reads local progress trackers (e.g., LEARNING.md) and fetches remote lesson markdown (zh.md) and quiz data (quiz.json) from raw.githubusercontent.com/fancyboi999/ai-engineering-from-scratch-zh/.
Boundary markers: There are no explicit delimiters or safety instructions provided to the agent to treat the fetched content purely as data rather than instructions.
Capability inventory: The skill has the authority to write to local progress files and execute shell commands to run lesson code snippets when a runtime environment is available.
Sanitization: The skill does not define any validation or sanitization routines for the content fetched from the remote repository or the local state files.- [EXTERNAL_DOWNLOADS]: The skill fetches educational materials (markdown and JSON) from the author's GitHub repository. These resources are hosted on raw.githubusercontent.com.- [COMMAND_EXECUTION]: During the "动手构建" (Hands-on construction) phase of the lessons, the skill instructs the agent to execute code segments from the repository and display the results to the learner.

learn