The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to execute a local Python script, atoms-search.py, to retrieve relevant knowledge snippets for course generation. The script utilizes standard library modules to perform keyword-based scoring and filtering on local or remote datasets.
[EXTERNAL_DOWNLOADS]: The search utility makes network requests to https://api.xyskill.xyz/v1/atoms/search to access a full repository of knowledge atoms. This domain corresponds to the vendor's known infrastructure and is used for non-executable data retrieval.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied content and local project files to dynamically generate course structures, creating a surface for potential instruction injection.
Ingestion points: User-provided text snippets, material uploads, and core project files such as SOURCE_OF_TRUTH.md and README.md (specified in Phase 1.5).
Boundary markers: The skill does not explicitly define delimiters to isolate external text from system instructions in its generation templates.
Capability inventory: The skill is capable of writing Markdown files to specific directories and executing local Python scripts to query knowledge databases.
Sanitization: The skill includes explicit instructions for the agent to bypass sensitive files and directories, including .env, .git/, .private/, API keys, and financial documents, which significantly mitigates the risk of accidental data exposure.

xy-course