Guides
Claude practitioner guides
These are working notes, not documentation. Each guide answers one question that comes up when you actually build with Claude — what to do when the context window fills, which MCP transport survives your deployment, how to stop an agent loop that will not finish — and takes a position instead of listing options.
-
Context & Memory
Compaction or truncation — what happens when context runs out?What compaction and truncation each do when a context window fills, how to tell which one broke your session, and which to pick for exact vs gist-level context.
Published 1 August 2026
-
Tools & MCP
How do you build and ship an MCP server?The build order for a Model Context Protocol server: primitives, transport choice, OAuth and input validation, tool design, and deployment.
Published 1 August 2026
-
Context & Memory
How do you hand off work to a fresh Claude session?A handoff template that lets a session with no history resume mid-task: state, decisions with reasons, dead ends, verbatim errors and one exact next step.
Published 1 August 2026
-
Agents & Orchestration
How do you stop an agent loop that won't finish?The four non-termination patterns, the six stop conditions every agent loop needs, and how to end a run with an actionable handoff instead of a silent timeout.
Published 1 August 2026
-
Reliability, Security & Trust
How do you stop an LLM from hallucinating?Classify the four hallucination types, then apply controls that are enforced in code: retrieval, verifiable citations, strict schemas and harmless failure paths.
Published 1 August 2026
-
Tools & MCP
stdio, SSE or Streamable HTTP — which MCP transport?How to choose an MCP transport by deployment reality — who runs the process, who holds credentials, who upgrades — plus the failure modes of each option.
Published 1 August 2026
-
Context & Memory
What is CLAUDE.md and what should you put in it?What CLAUDE.md is, where Claude Code looks for it, what belongs in it (and what never should), plus a working example and review checklist.
Published 1 August 2026
-
Agents & Orchestration
When is fanning out to subagents worth it?What parallel subagents actually buy you, what they cost in tokens and debuggability, and a four-part test for whether to fan out or stay sequential.
Published 1 August 2026
-
Prompting & Output Craft
Why does the output format drift, and how do you stop it?The six real causes of output format drift, why prompt pleas do not fix it, and the structural changes (schema, rejecting parser, repair metric) that do.
Published 1 August 2026
How the guides relate to the tracks
A track is a Socratic dialogue: the coach probes your reasoning and surfaces the concepts you fumbled. A guide is the same concept written down, for when you want the answer rather than the questions. Read the guide, then practise the decision with the coach.
Exam simulator — waitlist
A timed exam simulator is in build, not launched. If you want it when it ships, join the waitlist.