What does agent-friendly-cli do?
Use when designing, refactoring, or reviewing CLI tools and their accompanying Skills, especially those meant to be reliably invoked by AI agents while remaining usable by humans. Covers command surface design, non-interactive mode, JSON output contracts, exit code hierarchy, dry-run, idempotency, authentication security, responsibility layering of CLI/script/Skill/Agent, transactional orchestration, risk authorization, end-state verification, and acceptance methods.