← ClaudeAtlas

automation-designlisted

Design, simplify, and improve autonomous workflows, supervisors, scheduled jobs, and recovery loops across projects. Use when automation needs repeated human intervention, reports activity without results, or tries a new recovery approach. Record outcomes and refine this skill from evidence.
kookr-ai/kookr · ★ 5 · AI & Automation · score 70
Install: claude install-skill kookr-ai/kookr
# Automation Design Build automation that notices missing results, takes a useful authorized action, and checks its effect. Keep the control loop small enough to explain and test. Reliability means that failures are visible, bounded, and recoverable; no prompt can anticipate every edge case or supply missing authority. Use this skill to design or change the workflow. Use [[playbook-authoring]] for Kookr packaging and launch contracts, [[autonomous-watch-loop]] for a running watcher, and [[self-continuation-task]] for sequential work. Preserve the project's existing delivery and safety rules; this skill grants no additional permissions. ## Establish what success means Read the current workflow, authoritative state, and a few complete recent runs, including a failure and a healthy run when available. For an incident, reproduce the actual failed decision before designing its replacement. Read relevant [prior lessons](references/lessons.md); treat their evidence limits as part of the lesson. Define the smallest useful contract in the project's existing workflow document: - **Result:** the next observable outcome and the full completion criteria. Distinguish an independently useful slice from final activation or acceptance. - **Evidence:** where to verify that result, its identity or revision, and when it actually happened. Identify missing, stale, partial, and conflicting reads. - **Authority:** who owns the work; permitted corrections; actions that require outside in