{"id":53826,"plugin_id":"plugins_6ac5e934d1f88191b73fe472c16744ae","kind":"skill","collection_source":"plugin_package","comparison_source":null,"observed_at":"2026-10-24T13:55:10.323Z","digest":"278af452b293594ce8084fb95baccfe22fde25b2ecaa89b9cc24605c98068f45","against":null,"payload":{"description":"Assess process maturity and design responsible automation or AI support with human ownership, privacy controls, and explicit decision boundaries.","included_files":[],"name":"systems-ai-readiness","skill_md_contents":"---\nname: systems-ai-readiness\ndescription: Assess process maturity and design responsible automation or AI support with human ownership, privacy controls, and explicit decision boundaries.\n---\n# Systems and AI Readiness\n\nUse when a user wants to document, automate, or apply AI to a business process. A diagram or recommendation does not mean the system exists or has been implemented.\n\n## Readiness sequence\n\n1. Define the customer or operating outcome and the manual process that currently achieves it.\n2. Confirm that the process is real, stable enough to document, and has a named accountable owner.\n3. Map inputs, decisions, outputs, exceptions, handoffs, and failure consequences. Distinguish rules from judgment.\n4. Identify suitable low-risk administrative assistance. Keep high-impact judgment with qualified humans unless a separately validated and governed system is approved.\n5. Define data purpose, minimum necessary fields, provenance, access, notice/consent, retention, deletion, security, and vendor handling.\n6. Specify human review, override, correction, escalation, stop controls, logging, and responsibility for errors.\n7. Test realistic edge cases and failure modes before limited deployment; define ongoing monitoring and rollback criteria.\n8. Verify available tools and integrations before claiming technical capability. Do not invent connections, permissions, or performance.\n\n## Decision boundary\n\nAI should not independently make consequential safety, eligibility, diagnosis, legal, financial, hiring, or service-commitment decisions based solely on an unvalidated workflow. Require qualified human judgment, explicit controls, and appropriate domain review. Keep customer-facing actions human-approved unless the process has been separately authorized and governed.\n\n## Standard output\n\nProvide the current maturity and evidence status, process map, potential assistance points, excluded decisions, owner/reviewer, data controls, failure handling, validation plan, and a bounded recommendation such as **DOCUMENT MANUALLY FIRST**, **READY TO TEST A LIMITED ASSIST**, or **NOT READY**.\n"},"changes":[],"summary":"First saved snapshot. No earlier version is available for comparison.","summary_kind":"deterministic","summary_metadata":{}}