Everything free I've made for building with Claude Code and AI agents. No strings, no email required. I keep this same link updated, so save it.
Follow me on Instagram for new drops — @angelotrifanoff.ai
angelo-swe
angelo-swe
/ starter-kit.md
Last active
August 15, 2026 22:10
Free Claude Code Starter Kit — everything I've made, kept updated
angelo-swe
/ repo-series.md
Last active
August 29, 2026 03:58
The Repo List — every repo from my series, kept updated
Every repo from my series, newest first. I keep this same link updated — save it.
Follow me on Instagram for the next one: @angelotrifanoff.ai
https://github.com/MadsLorentzen/ai-job-search (MIT, ★36K+)
angelo-swe
/ claude-code-loop-cheatsheet.md
Created
July 21, 2026 00:58
The 4 AI Loops Cheat Sheet — stop prompting Claude Code one message at a time
Most people use an AI agent like a chatbot: prompt, wait, prompt again. Claude Code has four kinds of loops — each one hands off more of the babysitting. Here's all four, with a copy-paste example for each.
(Based on Anthropic's "Loop engineering: Getting started with loops" — claude.com/blog/getting-started-with-loops)
angelo-swe
/ CLAUDE.md
Created
July 20, 2026 01:54
The CLAUDE.md I make Claude Code read every session — file deletion safety, npm supply-chain rules, adversarial review, and the Karpathy guidelines
This is a sanitized copy of the real
CLAUDE.mdI keep at the root of my dev folder. Client work and internal project notes are stripped out — every rule below is one I actually run, learned the hard way, and re-read by an AI agent hundreds of times a week.
Always use trash instead of rm or rm -rf. The trash command (macos-trash) moves files to macOS Trash instead of permanently deleting them. If trash fails or isn't supported for a specific operation, let me know before falling back to rm.
angelo-swe
/ 72-hour-rule.md
Created
July 19, 2026 01:09
The 72-Hour Rule — stop your AI from installing poisoned packages (CLAUDE.md rule + terminal guard)
Why this exists: in May 2026 the "Mini Shai-Hulud" worm compromised 84 versions across 42 popular npm packages (the @tanstack family plus others) using malicious install scripts — code that runs the moment the package installs. Almost all supply-chain attacks like this get detected and pulled within the first 2–3 days. So the defense is patience, enforced.
angelo-swe
/ adversarial-review-rule.md
Created
July 18, 2026 19:38
The Adversarial Review rule — make a second AI break your AI's code (CLAUDE.md rule from @angelotrifanoff.ai)
angelo-swe
/ fablethink.md
Created
July 16, 2026 00:08
fablethink — a problem-solving doctrine for AI coding agents, modeled on how Anthropic's Fable-class models think. Save to ~/.claude/skills/fablethink/SKILL.md
| name | fablethink |
|---|---|
| description | Use when the user explicitly invokes /fablethink — or explicitly asks to "think like Fable" or apply "fablethink" — on a problem in any domain. Manual invocation only. Do not auto-load this skill, and do not apply it unless the user has explicitly asked for it. |
Problem-solving doctrine modeled on Anthropic's Fable-class models. Two failure modes destroy answer quality, and a fluent model is more susceptible to both because its first draft looks finished: converging too early (on a reading of the task, a diagnosis, or an approach) and trusting an answer never checked against reality. Everything below defends against one of those two — sized to the problem.