What problem does it solve? Building a reusable babysit/watch/retry loop normally requires hand-writing a loop instruction doc, a shell driver, and an Invoker workflow YAML, each with strict contracts that are easy to get wrong. This Skill turns one structured interview into all three artifacts with consistent naming, safety rules, and validation. ## Core Features & Use Cases - Interview-driven generation: Resolves a 15-field schema (goal, success criteria, evidence sources, write mode, submission mode, and more) before drafting anything, so generated loops are concrete rather than guessed. - Three coordinated artifacts: Produces a loop instruction markdown doc with a fixed section contract, a driver shell script with safe state-copy inspection and failure dedupe, and an Invoker workflow YAML routed through the plan-to-invoker handoff pipeline. - Benchmark/direct-output mode: For headless prompts with a literal required output path, writes the artifact directly without clarifying questions, validation, or submission. - Use Case: Ask for a worker-only PR babysitting loop; the Skill interviews you on success criteria and blockers, then generates the loop doc, driver script, and validated Invoker YAML ready for review and submission. ## Quick Start Ask the assistant to use the loop generator to build a reusable watch-and-retry loop and answer its interview questions about goal, success criteria, and write mode.
