The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to execute shell commands to manage files and directories. This includes using mkdir -p to build the hierarchy, git mv and git commit for version control, and pnpm ai-hero-cli internal lint to validate the created structure. These operations are consistent with the skill's stated purpose of scaffolding exercises.
[INDIRECT_PROMPT_INJECTION]: The skill relies on parsing a "plan" provided in the user context to determine what directories and files to create, which introduces a potential attack surface for indirect prompt injection.
Ingestion points: Untrusted data enters the agent context through the "plan" used in the parsing workflow defined in SKILL.md.
Boundary markers: The instructions lack explicit delimiters (e.g., XML tags or triple quotes) or warnings to the agent to ignore instructions embedded within the exercise plan.
Capability inventory: The skill possesses the ability to create directories, create/write files (readme.md, main.ts), move files via git, and execute a local CLI tool.
Sanitization: There is no mention of validating or escaping the plan's content before using it to generate directory names or file content.

scaffold-exercises