UNPKG

major-ai-skills

Version:

Installable agentic skills / AI agent skills (SKILL.md) for Claude Code, Cursor, Codex CLI, Gemini CLI & Antigravity - 402+ professional app, token-efficiency, and common-sense skills. SEO/GEO ready.

119 lines (90 loc) 5.16 kB
--- name: remember-ai-cutoff-dates description: "Check current sources for version-sensitive facts, APIs, prices, and rules before relying on model knowledge." category: common-sense risk: safe source: self source_type: self date_added: "2026-08-26" tags: ["knowledge-cutoff", "temporal-grounding", "api-deprecation", "versioning", "fact-checking", "prompt-engineering"] tools: ["claude", "cursor", "gemini", "codex", "chatgpt"] --- # Remember AI Knowledge Cutoff Dates (Temporal Grounding) (AI Skill) ## Overview Every foundational AI model operates with a fixed **Knowledge Cutoff Date** - a point in time beyond which its neural weights have zero training data. When queried about events, software releases, or pricing changes that occurred after its cutoff, offline models will not say *"I don't know"*; they will extrapolate older, obsolete information with total grammatical confidence. The **Temporal Grounding Protocol** ensures that time-sensitive queries are anchored to current release cycles, forcing models to state their cutoff or utilize live web search tools for real-time accuracy. --- ## Static Training Weights vs. Live Web Retrieval ``` ┌─────────────────────────────────────────────────────────────┐ Temporal Awareness Matrix [ STATIC WEIGHTS ONLY (Offline LLM) ] Excellent for timeless fundamentals: SQL, algorithms, calculus, standard biology, Latin roots, classic prose HIGH RISK: Modern framework syntax, cloud pricing, laws [ TEMPORAL GROUNDING / WEB RETRIEVAL TRIGGER ] Mandatory for: SDK releases, breaking API changes, fiscal year tax thresholds, live SaaS pricing tiers └─────────────────────────────────────────────────────────────┘ ``` --- ## High-Risk Temporal Vulnerability Zones 1. **JavaScript & Frontend Frameworks**: Next.js (Pages Router vs. App Router), React 18/19 Server Actions, Vite. 2. **Cloud & SaaS Pricing**: AWS instance pricing, OpenAI / Anthropic model rate limits and tiers. 3. **Tax & Legal Regulations**: Annual IRS standard deductions, 401(k) contribution caps, EU AI Act compliance. 4. **Corporate Leadership & Mergers**: Current CEOs, acquired startups, rebranded tools. --- ## Master Temporal Grounding Prompt Templates ### Pattern 1: The Explicit Version & Year Anchor (Coding & SDKs) ```markdown I am building a backend in [LANGUAGE / FRAMEWORK: e.g. Next.js 15 App Router / Python 3.12]. Temporal Rules: 1. Provide code compatible strictly with [FRAMEWORK VERSION]. 2. Do NOT use deprecated APIs or older patterns (e.g. do not use `getServerSideProps` or deprecated middleware conventions). 3. If your knowledge base does not cover the latest version release, explicitly state: `⚠️ [OUTDATED KNOWLEDGE CAVEAT: Specify cutoff date]`. ``` --- ### Pattern 2: The Live Search Enforcement Directive Use when using models equipped with web browsing capabilities: ```markdown What are the current pricing tiers and storage allowances for [SaaS PRODUCT: e.g. Supabase]? Requirement: - Use your live web search tool to check the official live pricing page today. - State the exact retrieval date and source URL. - Do not rely on training memory. ``` --- ## Real-World Case Study ### Scenario: Implementing Next.js Server Components #### Without Temporal Grounding (Deprecated Legacy Code) > **User**: *"How do I fetch data from a database in Next.js?"* > *Offline AI generates legacy `getServerSideProps` code from 2021 Pages Router architecture, which is deprecated in modern Next.js App Router.* #### With Temporal Grounding Anchor > **User**: *"Write a database fetch in Next.js 15 using Server Components and async/await."* **AI Output (Modern, Production-Ready)**: ```typescript // app/users/page.tsx (Next.js 15 Server Component) import { db } from "@/lib/db"; export default async function UsersPage() { // Direct server-side async query - zero API boilerplate const users = await db.user.findMany({ orderBy: { createdAt: "desc" }, }); return ( <main className="p-6"> <h1 className="text-2xl font-bold">Active Users</h1> <ul> {users.map((user) => ( <li key={user.id}>{user.email}</li> ))} </ul> </main> ); } ``` --- ## Summary Best Practices - **Always specify the year or version in code prompts**: Say *"Python 3.12 with Pydantic v2"*, never just *"Python"*. - **Look out for "Hallucinated Continuity"**: If an app rebranded recently (e.g., Twitter $\rightarrow$ X), specify the new name and API endpoints explicitly.