optimize-agentic-cli

Make a CLI an agent can drive without human babysitting. The contract: pure machine-readable stdout, semantic exit codes, non-interactive defaults, structured errors with retry guidance, and — for workflows that produce repairable artifacts — an iterative feedback loop the agent can finish on its own.

When To Use

Trigger this skill when the user is doing CLI-only agent-readiness work:

• "make my CLI agent-friendly" / "audit CLI for LLM use"
• the agent keeps failing to parse a CLI result
• stdout mixes JSON with progress text or spinners
• every failure exits 1 with no structured error body
• a command hangs waiting for an interactive prompt
• the workflow validates, rejects, repairs, and resubmits a generated artifact
• a fresh CLI needs agent-first constraints from day one
• help text is unreliable as a contract for an agent caller

Do not use this skill for: