{"id":77036,"plugin_id":"plugins_6ab4123730588191bc9267b9458b2670","kind":"skill","collection_source":"plugin_package","comparison_source":null,"observed_at":"2026-11-03T00:12:20.154Z","digest":"d3b8e2e52bcfd397a6d2475acf7caa947095b3d0f338884c1271620918fd3262","against":null,"payload":{"description":"Define when an AI feature is justified and how it should be scoped, evaluated and secured.","included_files":[],"name":"plan-ai-feature","skill_md_contents":"---\nname: plan-ai-feature\ndescription: Define when an AI feature is justified and how it should be scoped, evaluated and secured.\n---\n\n# Plan Ai Feature\n\n## When to use\nDefine when an AI feature is justified and how it should be scoped, evaluated and secured.\n\n## Workflow\nStart with user task and compare a deterministic workflow before adding a model. Assign narrowly bounded responsibilities to AI. Define input/output contracts, structured validation, source quality, fallback and human review. If retrieval or agents are considered, explain why and specify tools/authorization, injection boundaries, evaluation sets, latency and cost budgets.\n\nOutput AI feature contract, alternatives, failure taxonomy, evaluation strategy and what is excluded. Never promise autonomy or correct answers without evidence. Use `../../references/platform-product-patterns.md` and `../../references/nonfunctional-quality.md`.\n\n## Deliverable and truthfulness\nLead with the decision or result, then show supporting evidence, explicit assumptions, risks, and a concrete next step. Adapt depth to the user's ask. If required tools or source material are missing, disclose the limit; do not simulate inspection, implementation or research. Never write/modify/deploy without relevant authorization.\n\nRelevant reference: `../../references/platform-product-patterns.md`.\n"},"changes":[],"summary":"First saved snapshot. No earlier version is available for comparison.","summary_kind":"deterministic","summary_metadata":{}}