Real patterns, not theory
Every technique comes from production builds — skills, runners, and architecture decisions that were tested against actual projects before they were written down.
The QuassLabs Newsletter
AI architecture worth reading.
Practical patterns from a working AI engineering practice — skills, agents, local models, and the architecture decisions that ship products. Includes the free 19-skill starter kit.
No spam. No cadence quota. Issues ship when there is something worth saying. Unsubscribe any time.
What you get
One building practice, distilled into five things each issue delivers.
AI tooling that compounds
The newsletter covers the Claude Code skill system, local model routing, and agent patterns that grow more powerful the more you combine them.
Free kit with every issue
Subscribers get early access to new skill releases and kit updates before they hit the public download page.
No padding, no fluff
Issues ship when there is something worth shipping — a new pattern proven in a real project, a tool worth your time, or a hard lesson from a build gone sideways.
Practical AI architecture
Memory layers, orchestration patterns, model routing, and cost control — the architecture side of AI that most newsletters skip.
QuassLabs Free Starter Kit
Free download
19 production-tested Claude Code skills pulled from a private 400+ skill hub.
Each skill ships as a triad: SKILL.md, FUNCTIONS.md,
and LESSONS.md. Drop the folder into any Claude Code project and invoke by name.
MIT license. No account required.
- agent-setup-wizard Bootstrap any new project with CLAUDE.md, AGENTS.md, and context files
- skill-builder Draft a new skill with the correct triad structure
- skillmaster Full lifecycle manager — prior art, lint, sync, promote
- skill-linter Validate triad completeness and metadata consistency
- task-router Route tasks to the right model or agent based on complexity
- ollama-task-router Local-first inference routing (Ollama vs cloud)
- notify Telegram alerts for checkpoints, errors, and reports
- session-handover Generate a context-limit handover to HANDOVER.md
- reflect End-of-session retrospective to LESSONS.md
- memory-ladder Cross-session memory management (Layer 4+)
- test-loop-ug-ug-logger Ultra-compressed test execution logging (60-70% token savings)
- update-config Configure Claude Code settings.json and hooks
- keybindings-help Customize Claude Code keyboard shortcuts
- humanizer Remove AI-signature patterns from generated text
- deep-research Fan-out web research with adversarial claim verification
- code-review Multi-level diff review with optional inline PR comments
- simplify Reuse, efficiency, and altitude cleanups on changed code
- security-review Targeted security pass on a file or PR diff
- mermaid-prerender Pre-render Mermaid diagrams to inline SVG for file:// delivery
Each newsletter issue is published as a post on the blog. Browse the archive