The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands including mkdir, git mv, and git commit for repository management, alongside pnpm ai-hero-cli for linting. These are standard development operations for managing project structure and ensuring data consistency.
[PROMPT_INJECTION]: The skill processes a user-provided 'plan' to determine the directory structure, which constitutes an indirect prompt injection surface. Evidence: 1. Ingestion point: parsing the plan for section and exercise names (SKILL.md). 2. Boundary markers: none present in the instructions. 3. Capability inventory: filesystem directory creation, moving files, git commits, and local CLI execution. 4. Sanitization: none specified. The risk is assessed as safe because the operations are confined to standard repository scaffolding tasks.
[SAFE]: No evidence of data exfiltration, hardcoded credentials, or code obfuscation was detected during the analysis.

scaffold-exercises