Claude AI Resources: One Map for Videos, Repos, and Papers

Most "complete" link dumps are neither complete nor ordered. You get three near-duplicate intro videos, five skills repos that do the same thing, and papers listed because they look impressive—not because they help you ship. This page is smaller on purpose: one entry per niche where I could justify it, duplicates folded, and a suggested order if you are learning Claude seriously in 2026.

Who this is for: Engineers who already write code and want Claude Code, agents, or API workflows without wading through affiliate spam.

---

Videos (shortest path first)

Two entries below share the same YouTube ID (6eBSHbLKuN0)—different titles on the same upload. Watch once; do not bookmark both.

Mastering Claude Code in 30 Minutes (also titled "Claude Code Best Practices | Code w/ Claude")

Fast orientation to the terminal workflow—good first watch.

Claude FULL COURSE (~1 hour)

Broader build-and-automate framing if you want depth after the 30-minute pass.

36 Claude Tips for Beginners

Dense tips format; skim chapters in the progress bar if you are impatient.

Automate Any Task Using Claude (playlist)

Workflow-shaped; use when you care about end-to-end automation, not single prompts.

Ultimate Claude 4.1 Guide (2026)

Model-family overview—verify claims against current release notes; model names drift.

---

Repositories (clone in this order)

Repo	Why open it
anthropics/claude-code	Official Claude Code—source of truth for the CLI behavior you are debugging.
anthropics/claude-cookbooks	Copy-paste patterns for API and tool use; good when examples beat prose.
anthropics/claude-agent-sdk	When you outgrow ad-hoc scripts and want agent-shaped code.
obra/superpowers	Opinionated workflows (planning, review loops)—adopt piecemeal, not wholesale.
travisvn/awesome-claude-skills	Directory of community skills; treat as a menu, not a checklist to install all.
travisvn/claude-d3js-skill	Niche: charts and D3—only if your product actually ships visualizations.
travisvn/loki-mode	Heavy orchestration experiments—read the README before running anything on main.
yusufkaraaslan/Skill_Seekers	Tooling to turn docs into skills—useful if you maintain internal knowledge bases.

---

Official guides and PDFs

• Claude prompting best practices — API- and product-oriented; read before you tune system prompts for production.
• Claude Code best practices — terminal workflow, permissions, and repo hygiene.
• The Complete Guide to Building Skills for Claude (PDF) — skills architecture on Anthropic’s terms; keep a local copy—URLs to PDFs move.
• Anthropic’s constitution — not a tutorial; context for how safety rules show up in product behavior.
• Claude Insider documentation — third-party; cross-check anything that touches billing or account security against official docs.

---

Books (general AI engineering, not Claude-only)

Claude changes monthly; books stay static. Use them for systems thinking, not for "current model name" facts.

• Mastering Claude AI: Practical Journey from First Prompts to Pro — prompt-to-workflow progression; verify anything version-specific against docs.
• Claude LLM by Anthropic (audiobook) — audio-first learners only; same caveat on freshness.
• AI Engineering (Chip Huyen) — production patterns for models and data—not Claude-specific, but it is the right layer.
• Designing Machine Learning Systems (Chip Huyen) — when your "Claude project" is really an ML system with evals and monitoring.

I dropped a fifth title that reused the same Amazon ASIN as another listing—likely a catalog error—rather than pass bad metadata along.

---

Papers (why they matter for agents)

You do not need arXiv to use Claude. You do need these if you argue about tool use, memory, or multi-step reasoning in design reviews.

• ReAct — interleaving reasoning and actions; baseline for "agent loops."
• Generative Agents — memory and simulation; read before you build "persistent personas."
• Toolformer — when to call external tools from language models.
• Chain-of-Thought Prompting — still the reference for explicit intermediate steps.
• Tree of Thoughts — search over reasoning paths; expensive, know the cost before shipping.
• Reflexion — verbal reinforcement-style self-critique; pairs with eval harnesses.

---

Trade-offs you should actually care about

• More skills ≠ better outcomes. Each skill is context and maintenance—ship three that are tested.
• Videos age fast. Prefer official docs for flags, model IDs, and pricing.
• Community repos vary in quality. Read issues, check last commit date, fork before you depend on it in prod.

---

A minimal weekend path

1. Watch the 30-minute Claude Code video once.
2. Clone claude-code and skim the README for your OS.
3. Read the Claude Code best practices doc end-to-end.
4. Pick one paper (ReAct or Toolformer) and skim the abstract and figure 1.

After that, your bottleneck is usually evals and permissions—not more links. If something here 404s, search the Anthropic docs site and open an issue on our side; we will rotate the list.