The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill parses user-supplied plan text to determine directory and file names for scaffolding.
Ingestion points: Natural language plans containing section and exercise names provided by the user in the prompt flow.
Boundary markers: No specific delimiters are used to wrap the input plan to prevent instructions embedded within the plan from being misinterpreted.
Capability inventory: Filesystem modifications using mkdir -p and git mv, as well as execution of pnpm ai-hero-cli for linting.
Sanitization: The instructions do not specify a requirement for the agent to sanitize or escape characters in the plan text before using them in shell commands.
[COMMAND_EXECUTION]: The skill executes local shell commands to automate the course scaffolding workflow.
Evidence: The skill workflow explicitly instructs the agent to run mkdir -p, git mv, and pnpm ai-hero-cli internal lint to set up and validate the directory structure.

scaffold-exercises