The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses several shell commands to manage files and validate structures.
Evidence: Use of mkdir -p to create nested directories, git mv for renaming, and pnpm ai-hero-cli internal lint to validate the resulting structure.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow where user-provided data (the 'plan') directly influences filesystem operations without explicit sanitization instructions.
Ingestion points: Workflow Step 1 in SKILL.md ('Parse the plan
extract section names, exercise names...') ingests untrusted user data into the agent's context.
Boundary markers: Absent. There are no instructions to the agent to treat plan content as potentially malicious or to strictly delimit it.
Capability inventory: The agent has the capability to create directories (mkdir -p), move files (git mv), and execute a project-specific CLI (pnpm ai-hero-cli).
Sanitization: Absent. The instructions do not mandate validation or escaping of the directory names extracted from the user's plan, which could be exploited for directory traversal (e.g., using ../ in a section name).

scaffold-exercises