The Agent Skills Directory

[COMMAND_EXECUTION]: The helper script scripts/publish_ai_course.py uses subprocess.run to execute git commands for repository metadata extraction (remote URL, branch name, status). These calls use a fixed list of arguments and are performed within the local course directory.
[COMMAND_EXECUTION]: The SKILL.md file contains workflow instructions using shell commands such as find, git, npm, and node. These commands are used for file inspection, website synchronization, and build verification within the user's development environment.
[PROMPT_INJECTION]: The skill processes untrusted data by extracting titles and metadata from Markdown files in the course source. This data is used to populate website configuration files, representing an indirect prompt injection surface. [Ingestion points]: Reads README.md, introduction.md, and syllabus.md from the user-provided course root. [Boundary markers]: None; data is processed as plain text and interpolated into YAML structures. [Capability inventory]: File system writes via write_yaml and shell execution via npm scripts and node commands. [Sanitization]: Validates the course_id against a safe regex and uses yaml.safe_dump to ensure valid formatting of the output configuration.

makecourse