The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to parse and act upon a user-provided plan, which introduces a surface for indirect prompt injection where malicious instructions could be embedded in the plan data.
Ingestion points: The user-supplied course plan described in the Workflow section of SKILL.md.
Boundary markers: None specified to differentiate between data and instructions within the plan.
Capability inventory: Filesystem creation (mkdir), file writing (readme stubs), and execution of local tools (pnpm, git) as defined in SKILL.md.
Sanitization: The instructions recommend dash-case naming for directories but do not provide explicit logic for sanitizing or validating user-provided plan content.
[COMMAND_EXECUTION]: The skill executes local command-line tools for linting and version control during its normal workflow.
Evidence: The skill invokes pnpm ai-hero-cli internal lint to validate the directory structure and uses git commit and git mv for file management in SKILL.md.

scaffold-exercises