Tactical Agentic Coding

8 Tactics

The mission: transform from a phase-one AI coder into a phase-two agentic engineer, the kind they cannot replace, by turning your codebase into a self-operating machine.

Tactical Agentic Coding is eight lessons, each built around a single tactic. Imagine eight tactics stacked on top of each other; you break through them one by one. Some are required, some accelerate your capabilities, and all of them compress into the final one: prioritize your agentic layer.

The core idea underneath every tactic is the Agentic Prompt: a plain prompt that reliably executes long chains of tool calls, wrapped in a programmable agentic coding tool that runs in the terminal. That capability upgrades the big three of AI coding (context, model, prompt) into the Core Four by adding Tools.

The eight tactics

#	Tactic	What it unlocks
1	Stop coding	Use the best tool for the job. Reallocate your cycles to planning, reviewing, and building loops.
2	Adopt your agent's perspective	The 12 Leverage Points and the 4 KPIs. Give the agent what you would need.
3	Template your engineering	Encode problem-solving into reusable prompts, meta prompts, and plans.
4	Stay out the loop	AFK agents and the PETER framework. Prompt to PR.
5	Always add feedback loops	Closed-loop prompts: request, validate, resolve. Self-validating agents.
6	One agent, one prompt, one purpose	Focused context windows. Review velocity. Evals for the agentic layer.
7	Target Zero Touch Engineering	The North Star. Drop the human review for classes of work you trust.
8	Prioritize agentics	The meta-tactic. Spend at least half your time on the agentic layer.

The compressed software development lifecycle

The whole course scaffolds against a five-step lifecycle, with the goal of augmenting and then automating each step so your agents ship on your behalf.

flowchart LR
  P[Plan] --> B[Build / Code]
  B --> T[Test]
  T --> R[Review]
  R --> D[Document]
  D -.conditional docs.-> P

The secret Dan reveals in Lesson 7: it is not actually about the lifecycle at all. It is about composable agentic primitives you can arrange to solve any engineering problem class. The lifecycle is just a familiar package for delivering that idea.