AI Agent Practical Guides & Workflows

Battle-tested insights from a big tech engineer: tooling division, 5 mental shortcut prompts, agent-friendly E2E testing harnesses, and high-risk code review practices.

Clean writing without losing soul. Master the two core pillars: delete rigid AI tropes and inject living human voice. Complete breakdown of 10 patterns to cut, 5 rules for natural prose, and a 3-pass rewrite loop.

Demystifying AI Agent engineering: Context, Harness, Loop, Eval, and Graph. At the end of the day, it all comes back to prompts. Code guards boundaries, prompts guide reasoning.

Real enterprise AI delivery beyond demo: data governance, permission isolation, SOW boundaries, dual test-sets, and acceptance checklist.

Master the complete Codex workflow: 4 entry points, 7-step loop, 4-field structured prompt, 3-tier permission matrix, AGENTS.md rules, and 10 common pitfalls.

Execution success is not task completion! Master the 5-step UX closed-loop (Entry → Comprehend → Action → Feedback → Verify) across AI engineering, Artifacts, Web, Apps, and writing.

From concept breakdown, hardware checks, Ollama setup to 100% GPU acceleration verification and local API integration with interactive VRAM calculator.

Eliminate AI hallucination & context drift! Spec-Driven Development (SDD) masterclass covering progressive disclosure, AGENTS.md router and open-source agents-spec tooling.

How to systematically master Hugging Face? From Spaces web testing, Models quantization picks, Datasets sampling, to LoRA stylization and Collections archiving with 23 workflow screenshots.

How to build AI products from finding demands, setting boundaries, rapid MVPs, to complete redesigns? Train yourself like an evolving foundation model with 18 high-res diagrams and interactive boundary validators.

Dense vs MoE, Base vs Instruct vs VL, BF16 vs FP8 vs 4bit, GGUF vs Safetensors, Ollama vs llama.cpp vs MLX: A 5-layer architectural walkthrough with an interactive VRAM & quantization calculator.

Keynote authorities on stage, but treated as scapegoated contractors on site? A deep field survey across big tech, AI startups, and independent OPCs: identity dilemmas, internal friction, price wars ($27k vs $280 bids), and historical parallels with 1960s systems analysts.

Where are skills stored? Why do they disappear in new projects? Where does that long installed list come from? A comprehensive architectural guide: agentskills.io standard, dual scopes (User vs Project), 5 disk sources, and lifecycle protection (agent_created and userModified).

What is the true difference between LLMs and AI Agents? A comprehensive breakdown from token generation, tools & APIs, ReAct agent loops, terminal sandboxes to MCP protocols and Codex desktop. Includes a complete real-world guide to building a medical paper research agent with PubMed integration, AGENTS.md rules, and reusable Skills.

Comprehensive illustrated manual for Tencent's native AI Agent workspace. 23 workflow screenshots detailing 3-panel UI, input operators (+/@//), context management, 10+ MCP connectors (Tencent Docs, QQ Mail, Canva, Notion), and 30k-row KOL data cleaning.

Comprehensive beginner guide to terminal agents. Setting up isolated demo folders, global installation, 4 UI zones, 4-element prompt engineering, and 5-point independent CLI verification.

Comprehensive practical guide to ByteDance's Doubao Work desktop agent. Unifying TRAE, Coze, and Lark, auto-aggregating 200+ local Codex Skills, and enabling Office automation, Lark Base sync, and Seedream 5.0 multimodal creation.

Struggling with Agent choosing wrong skills as list grows? Learn to fix it with descriptions, hierarchical routing, negative samples, and reranking.

End-to-end real case breakdown. 5-stage pipeline: feasibility solution matrix, hybrid design specs, baoyu-design high-fi UI prototyping, Claude Code /goal execution, and black-box QA testing without reading source code.

304-page full comprehensive book. In-depth breakdown of modern Agent architectures (LLM + Context + Tools), observation and action spaces, multi-agent collaboration, design paradigms, and engineering best practices.

Decoupling LLM brains from physical execution systems. Breaking down Agent ≈ LLM + Harness, comparing WorkBuddy, Codex, and DeepSeek Harness pluggable architecture.

Eliminating AI sycophancy and safe platitudes. Dual Steelmanning constructs the strongest analytical cases for both sides to expose the core divergence and decisive variables.

From installation, 3 specialized work modes, and Auto model routing, to custom third-party API integration, workspace isolation, and security confirmation workflows.

Feeling overwhelmed by 240+ frameworks? Deep-dive into Track A (CLI Power User) vs Track B (Agent Builder), featuring a 10-week step-by-step roadmap for everyday practitioners.

Lost in 70,000+ skills? Curated 15 hardcore skills across office productivity, coding, and social media matrix creation, with explicit trigger rules and context-slimming techniques.

Knowledge bases scale from simple PDFs to enterprise IAM and audit logs. Deep-dive into ad-hoc docs, Feishu spaces, Obsidian graphs, LLM Wiki maintenance, and enterprise RAG/FDE engineering.

Tired of manual content repurposing? Learn how to distill your distribution workflows into structured Prompts to automatically generate 3 platform-tailored viral posts for LinkedIn and X.

Tired of manual RED content production? Learn how to use Codex to maintain your topic pool, title library, copywriting templates, and data analytics, converting repetitive tasks into a standardized project workflow.

Just saying 'Teach me Python' results in flat general answers? Learn to construct the 'Role→Method→Situation→Requirements' framework to turn AI into a deliberate practice coach.

An in-depth analysis on building adaptive context and memory inheritance with Codex, enabling complex agent development to continuously evolve beyond single prompts.