claude-agent-sdk

Note: While the most well kept content lives under a Claude directory, most rules etc. are portal between agentic coding tools, the exception being Claude _Skills_, which I'm yet to see other tools support - but have no doubt they will as they're one of the most powerful and useful agentic coding concepts.

• [Agent Rules](./Claude/CLAUDE.md)
• [Agent Skills](./Claude/skills/)
• [Agents / Sub-Agents](./Claude/agents/)
• [Agent Commands (Prompt Templates)](./Claude/commands/)

Client Tooling - Not All Created Equal

I _highly_ recommend Claude Code or Cline as the best agentic coding tools.

I find both Claude Code and Cline _far_ outperform the many other tools I've tried including Copilot Agent, Gemini CLI, Codex, Cursor (probably one of the lowest performing tools I've used down at the bottom with with Amazon Kiro), Kiro, Windsurf, Augment Code, OpenCode, Aider and many others over the past 3 years.

Claude Code and Cline are both good in different ways. Claude Code provides the best value for money if you can pair it with the Claude Max 5 subscription ($100 USD / month).

Since around May 2025 I've been using Claude Code far more any other tool:

• I've had a number of issues with both VSCode and VSCode fork based agents that annoy me. Stealing focus from the editor, opening panels on IDE launch, UI and state glitches etc. and generally have found moving to TUI based tooling more efficient for my workflow.
• [Claude Skills](https://www.claude.com/blog/skills) are a simple, highly effective way to extend coding agents with new knowledge and capabilities dynamically without bloating the context.
• I delegate work to [Claude Sub-Agents](https://code.claude.com/docs/en/sub-agents) all the time to vastly reduce context bloat and parallelise tasks.
• The Claude Max 5 subscription ($100USD+GST/mo) provides me with _far_ better value for money than what I would have to spend with consumption based pricing ($2000USD+GST/mo).

The only real down side to Claude Code is that it's not open source, which is a shame as I believe open source always wins in the long run.

When setup correctly you can do fantastic work in either Claude Code or Cline (and the Cline developers are a fantastic bunch!), but beware of other tools (especially those like Cursor or Kiro with large marketing departments and low monthly subscription costs).

I have a table where I'm trying to keep track of coding agent tools here: https://smcleod.net/agentic-coding-tools/

I've shared (somewhat out of date) overview of workflow (Setup -> Plan -> Act -> Review & Iterate) in a blog post [here (smcleod.net)](https://smcleod.net/2025/04/my-plan-document-act-review-flow-for-agentic-software-development/).

The TLDR is that this is a lightweight approach of having coding agents help me design a documented plan with the goal, context, requirements / constraints and a phased checklist of tasks to complete and have them update the plan as they work through the tasks in each phase.

Note: I've tried '_Spec Driven Development_' but find it often leads to over-engineering and heavy documentation debt.


For 99% of my work I only use a single MCP server: [MCP DevTools](https://github.com/sammcj/mcp-devtools)

I wrote it to provide the most common tooling I use with Agentic Coding.

• Always have tools available to the agent that allow it to:

- Search the web.

- Efficiently retrieve web page content as markdown.

- Lookup package documentation.

- Perform math calculations.

• Be mindful of how many tokens each MCP server adds to your context window (/context in Claude Code), some tools abuse their descriptions and pollute the context window - for example Github's official MCP server alone uses ¼ of the entire context window of Claude Sonnet 4.5.

---

Writing Rules

• Using pseudo-XML for your rules can help LLM adherence

- LLMs are trained on a lot of structured data (such as XML), see [this blog post](https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/use-xml-tags) for more information.

- On _many_ occasions I have been given feedback that after simply changing rules XML like structure - the LLM closer adheres to them.

• Only enable the rules you actually want to use

- Think about the signal to noise ratio of your rules (and context in general), _how much information could you be told at once and remember?_

- If you have a lot of rules that are not relevant to the current task, then you're just adding noise and misleading the prediction engine, while increasing the token count.

- Do not blindly import and enable all my (or anyone else's) rules!

• Be clear, concise and specific in your rules. Avoid ambiguity
• Use emphasis (e.g. bold, italic, __underline__) to highlight important parts of your rules.
• As well as global rules, consider adding project specific rules such as .clinerules, CLAUDE.md or similar that relate to repository specific behaviour (e.g. "To build the application, you must run make build etc.)

- I have an _example_ of what these might look like in [Cline/Rules/adhoc/_repo-specific-rules.md](./Cline/Rules/adhoc/_repo-specific-rules.md).

• Rules are often transferable between agentic coding tools

- While I write a lot of my rules in Cline, for 95% of them there's no reason they can't be used with other Agentic Coding tools such as Claude Code etc. without modification.

• Get AI to help you write or improve rules

If you spend a long time on a difficult problem with a coding agent and you finally crack it - get it to:

1. Summarise the fix
2. Why previous attempts did not work
3. What led them down the wrong paths initially
4. Get them to write a concise, clear rule (prompt) that could be used in the future (or added to your global rules if it's a common issue) to prevent the issue from happening again or at least aid with debugging.

Example:

> You fixed it! That's taken a long time to fix. Can you please respond with details on:

> 1. What the fix was

> 2. Why it wasn't picked up earlier

> 3. What information could I have provided to AI coding agents in the future - not just for this project but also other projects in general?

> With those in mind I would like you also like you to create a 1 to 3 sentence prompt I can provide to future AI coding agents that would help them avoid having similar issues in the future.

---

Getting High Quality Outcomes

• Treat an agent like you would someone who just joined your team, don't assume they know anything about your codebase or intended outcomes. Unless the task is very simple and self explanatory - a single sentence is probably not going to be enough for a prompt. GIGO.
• Manage the context window usage effectively (see other notes here on this).
• Start with a plan - break down large or complex tasks into a checklist of items to complete, have the agent follow and mark off items has it completes them.
• Make use of tools (MCP servers), they extend and enhance LLMs with access to up to date information, new capabilities and integrations.
• Always have tools available to the agent that allow it to search the web, lookup package doc