kcvs2002/ai-task-playbook
Task clarification, evidence-based verification, and reproducible debugging workflows.
Clarify an unfamiliar task from its supplied README, data, code, and stakeholder answers. Use at task start or when requirements conflict. Produce a brief, prioritized questions, and observable acceptance criteria; do not invent business requirements or implement before scope is understood.
Diagnose a reproducible error or repeated failed implementation in a local task. Use when execution fails, calculated results differ, or a service does not respond. Isolate the cause, make a scoped correction, and verify with the same reproduction; do not alter platform controls.
Verify an implemented analysis, script, API, web interface, or document against concrete requirements. Use after a meaningful implementation or before submission. Compare expected and observed results, record failures and unverified items, and rerun the same check after fixes.