| 030-model-selection |
base |
important |
Choose the right model — results depend on this heavily. Short answer: Claude Sonnet 4.6 |
| 035-visual-context-screenshots |
base |
|
Not just text — images can also be part of your context |
| 040-agent-mode-under-the-hood |
base |
important |
How the agent works under the hood — removes the magic |
| 050-effective-prompting-without-arguing |
base |
important |
Don't argue with models — they hallucinate more under pressure. Explore alternatives to arguing |
| 055-clarifying-requirements-before-start |
base |
important |
How to fill the context before starting — switching the agent into interview mode |
| 057-agent-memory-management |
base |
|
Offload any memory to markdown files — makes the model less forgetful |
| 058-workspace-kickoff-iterative-prompt |
base |
optional |
Create a kickoff prompt file that grows with UPD blocks — the iterative prompt approach for evolving investigations |
| 060-version-control-git |
base |
important |
Baby steps approach — critical when working with models that can break what was just done well |
| 070-custom-instructions |
base |
important |
FOUNDATION: Create instructions for any purpose. The base instruction for creating instructions — start every GenAI project here |
| 080-learning-from-hallucinations |
base |
important |
FOUNDATION: Don't edit instructions manually. When a hallucination occurs, ask the agent to fix the instruction so it won't happen again. A good instruction takes 10–15–20 iterations |
| 090-ai-skills-tools-creation |
base |
important |
FOUNDATION: The model can't do everything well — move part of the logic into a deterministic script, add an instruction, and get a Skill. Code is written by the model itself |
| 100-mcp-model-context-protocol |
base |
important |
A way to connect to 3rd-party systems |
| 103-cli-command-line-interface |
base |
important |
Another way to connect to 3rd-party systems |
| 110-development-environment-setup |
advanced |
|
For building your own PoC on Node.js — SpecKit can be wired up to it for rapid prototyping |
| 120-rapid-poc-prototyping-with-speckit |
advanced |
|
Spec-driven development framework — features are built better and the agent works autonomously longer |
| 130-chrome-devtools-mcp-qa-emulation |
advanced |
|
For web applications — how to automate QA in the browser |
| 160-bulk-file-processing-with-ai |
advanced |
|
When processing many files, write a script and work with the agent in CLI mode |
| 196-reverse-engineering-project-knowledge |
advanced |
|
When you have a project and need to extract instructions from its commit history |