Agentic Coding Path

• From Chatbot to Agent
• The Plan-Act-Observe Loop
• When the Loop Goes Wrong
• Knowing When to Step In
• What Makes a Good Agent Task
• The Manager Mindset

You already use AI for chat and autocomplete.

• Why Terminal Agents
• How to Evaluate Any Terminal Agent
• The Big Three in Practice
• The Broader Ecosystem
• Reading, Steering, and Recovering
• Daily Workflow

Evaluate, use, and steer any terminal agent.

• Why Agents Keep Getting It Wrong
• Writing Project Rules
• Shared Rules and Scoped Rules
• Cross-Session Context
• Documentation Agents Can Navigate
• Setting Up and Retrofitting
• Measuring Configuration Quality

Agents make the same mistakes until you tell them not to.

• Why Tasks Fail
• The Anatomy of a Good Task
• Break Features Into Pieces
• Self-Verifying Tasks
• Specs vs Step-by-Step
• When Tasks Go Wrong
• The Review Loop

Project rules tell agents how to work.

• The Agent Review Problem
• What Agents Get Wrong
• Reviewing Efficiently
• Calibrating Trust
• Drift in the Diff
• The Feedback Loop
• Getting Faster Without Getting Sloppy

Agents make mistakes in predictable patterns.

• When "Done" Isn't Done
• Recognize the Failure Fast
• Kill, Revert, Restart
• Save the Good Work
• Debug Code You Didn't Write
• Build Workflows Where Failure Stays Cheap
• Turn Failures Into Better Config

Agents break things in predictable patterns.

• Why One Big Prompt Fails
• Decompose Work Into Stages
• Sequence, Checkpoint, Commit
• When to Delegate
• When to Parallelize
• When the Pipeline Breaks
• Design the Workflow

Real features are too big for one agent session.

• When One Agent Isn't Enough
• Choosing the Right Coordination Mode
• Coordination Patterns
• Contracts, Ownership, and Handoffs
• Reviewing and Merging Multi-Agent Output
• When Coordination Breaks
• The Coordination Decision

One agent hits a wall when features span too many files.

• Reusable Agent Instructions
• Building the First Skill
• Controlling When Skills Activate
• Dynamic Context and Isolation
• Skills for Real Developer Work
• When Skills Misfire
• Sharing and the Configuration Stack

You repeat the same instructions every session.

• The Risk Framework
• Scoping Agent Access
• Keeping Secrets Safe
• Automated Safety Gates
• Audit Trails
• Emerging Safety Risks

Agents with broad file access and shell permissions can cause real damage without malice.

• What MCP Changes
• Connecting MCP Servers
• Building a Tool Ecosystem
• Custom MCP Servers
• MCP-Specific Security
• Debugging and the Complete Stack

Agents can only do what their tools allow.

• Why Agents Love Refactoring
• Spotting Agent-Friendly Debt
• Planning Refactors in Batches
• Rename, Restructure, Reorganize
• Making Legacy Code Agent-Ready
• Upgrading Dependencies Safely
• Multi-Session Refactoring
• Building a Refactoring System

Technical debt is where agents earn their keep.

• Choose Your Module
• Plan and Execute
• Verify and Merge
• Document and Repeat

Pick a module in a real codebase that needs work.

• Agents Beyond the Terminal
• Headless Execution
• Security Boundaries
• Read-Only Automations
• Write-Back Automations
• Debugging and Cost Control
• The Trust Progression

Agents don't have to wait for you to start them.

• When Size Becomes the Problem
• Context Selection
• Making Code Agent-Navigable
• Monorepo and Package Boundaries
• Multi-Repo Coordination
• Decomposition at Scale
• Scaling Agent Strategies

Large codebases break naive agent workflows.

• From Techniques to System
• The Delegation Framework
• Session Design
• Building and Evolving Patterns
• The Feedback Loop
• Team Adoption
• Your Workflow Document

You know the techniques.