Claude Code Skills: Build Reusable Workflows
Claude code skills guide: understand, choose, and apply Claude Code skills, with official sources, explicit controls, and reproducible verification.
Independent technical publication / Updated continuously
Field notes for building with AI agents.
Reproducible tests, code, versioned guides, and honest trade-offs for coding tools, MCP, evals, and the systems behind them.
LAB REGISTER / LIVEPRIORITY INDEX
Primary sourcesVersioned methodsFailure modes includedClear update history
01 / Priority deskThe highest-demand work,
The highest-demand work,
one click away.
Strategic launch pages sit directly on the homepage. Search signals are directional Semrush US estimates, not claimed traffic.
Codex CLI: The Evidence-Led Field Guide
Codex cli guide: understand, install, and operate Codex from a terminal, with official sources, explicit controls, and reproducible verification.
MCP Servers: Choose, Verify, and Connect
Choose MCP servers by task fit, maintainer evidence, access scope and failure behavior—not by directory popularity.
Claude Code: The Evidence-Led Field Guide
Claude code guide with official sources, explicit controls, tested failure paths, and a reproducible verification record.
Agentic AI: Definition, Boundaries, and Examples
Learn how to understand what makes a system agentic and what does not. Evidence artifact, failure checks, verification, and decision rule included.
Vibe Coding: Definition, Workflow, and Hard Limits
Learn how to understand where vibe coding helps and where it becomes unsafe. Evidence artifact, failure checks, verification, and decision rule included.
Install Claude Code on macOS, Linux, and Windows
Install claude code guide with official sources, explicit controls, tested failure paths, and a reproducible verification record.
Model Context Protocol: Architecture, Servers, and Risk
Understand MCP roles, primitives, transports and trust boundaries before choosing a server or writing an integration.
Claude Agent SDK: Build a Controlled Coding Agent
Claude agent sdk guide: decide when to move from the CLI to the Claude Agent SDK, with official sources, explicit controls, and reproducible verification.
Claude Code Agents: Subagents, Scope, and Control
Claude code agents guide: decide when and how to delegate work to Claude Code agents, with official sources, explicit controls, and reproducible verification.
Claude Code vs Codex CLI: Same Task, Same Repository
Claude code vs codex guide with official sources, explicit controls, tested failure paths, and a reproducible verification record.
Claude Code Hooks: Automate Without Losing Control
Claude code hooks guide with official sources, explicit controls, tested failure paths, and a reproducible verification record.
AI Code Review: A Verification-First Workflow
Learn how to use AI review without outsourcing engineering judgment. Evidence artifact, failure checks, verification, and decision rule included.
AI Agent Frameworks: A Tested Selection Guide
Choose an AI agent framework from a fixed production job, state model, control needs and operating evidence.
02 / Research channelsFollow the system,
Follow the system,
not the feed.
Each channel owns a distinct reader job and links its guides, labs and comparisons into one crawlable research path.
01Claude CodeSkills, hooks, subagents, context and reproducible workflows for Claude Code.8 live notes02Codex CLICodex setup, AGENTS.md, skills, sandboxing and grounded comparisons with other coding agents.5 live notes03Model Context ProtocolMCP architecture, servers, transports, security boundaries, integrations and debugging.6 live notes04Agent FrameworksFramework patterns, platform trade-offs and implementation paths for stateful agent systems.3 live notes05LLM Evals & ReliabilityEvaluation design, error analysis, datasets and repeatable checks for agent behavior.4 live notes06AI Code ReviewReview tools, false-positive control, pull-request workflows and evaluation methods.5 live notes07AI Coding AgentsOperating models, security limits, benchmarks and agent loops inside real repositories.1 live note08Agentic EngineeringContext engineering, observability and production patterns for teams shipping agent systems.3 live notes09Vibe CodingA sober look at fast AI-assisted building, its workflows, tools, checks and failure modes.1 live note
03 / Latest labnotesNew observations,
Browse the full index New observations,
kept in context.
Jira MCP: Setup and Permission Boundaries
Connect Jira through Atlassian Rovo MCP with bounded read, write and search permissions and direct record checks.
AI Code Checker vs Repository-Aware Review
Learn how to choose between pasted-code checking and repository-aware review. Evidence artifact, failure checks, verification, and decision rule included.
Langfuse Evals: Connect Traces to Tests
Learn how to derive and run evaluations from Langfuse traces. Evidence artifact, failure checks, verification, and decision rule included.
Context7 MCP: Documentation Context Under Test
Install Context7 MCP for one documentation task, verify source identity and compare retrieved context with official docs.
CodeRabbit Review: Setup, Test, and Limits
Learn how to test CodeRabbit on a representative pull request. Evidence artifact, failure checks, verification, and decision rule included.
DeepEval: Build a Reproducible Eval Suite
Build a small DeepEval suite from explicit cases and metrics, then pin, run and inspect it as normal test code.
04 / Editorial method
Every note has a visible way back to the evidence.
- 01Trace the source
Prefer specifications, repositories, papers and documented tests over recycled claims.
- 02Expose the method
Pin versions, constraints and the check that makes the result reproducible.
- 03Keep the limit
Publish uncertainty and failure modes beside the recommendation.