Claude: Settings & Config

A detailed 6-step system for building an Obsidian vault that auto-captures content via Readwise, Airr, Whisper, and a Telegram bot, routes everything through N8N pipelines, and uses Claude to generate daily briefings and weekly syntheses. The core thesis: knowledge systems fail because they optimize for input, not output—feedback loops (daily briefs, weekly synthesis) are what turn a dead archive into a thinking partner.

gstack is Garry Tan's open-source Claude Code setup that transforms a solo AI assistant into a structured 6-role virtual engineering team (CEO, Eng Manager, Designer, QA, Release Engineer, Doc Engineer) via 23 slash-command skills. Tan claims ~810× his 2013 productivity rate measured in normalized logical lines of code, shipping 3 production services and 40+ features in 60 days part-time. The system enforces a think→plan→build→review→test→ship→reflect sprint discipline where each skill feeds context into the next, and supports 10-15 parallel sprints via Conductor.

A GitHub-based plugin marketplace pattern allows centralized management of multiple Claude plugins with sub-skills and agents across an entire organization. Updates to the central repo automatically sync to all users' Claude instances, eliminating manual distribution overhead.

Hannah Stulberg (DoorDash PM) built a shared git repo where every team function checks in context—call summaries, decision logs, analytics queries—queryable in natural language via Claude Code. Four independent implementations (DoorDash, Pendo, Google, solo) converged on the same 3-layer architecture, suggesting a generalizable pattern for solving institutional knowledge loss.

Stop Slop is an open-source skill file (SKILL.md + reference files) that teaches Claude or any LLM to detect and remove characteristic AI writing patterns — banned phrases, structural clichés, and sentence-level rules — and score prose on five dimensions.

Opus 4.8 launches at the same price as 4.7 with new Dynamic Workflow mode (thousands of subagents for complex tasks), Ultracode mode (xhigh thinking + auto Dynamic Workflows), and a revamped fast mode on $200 plans offering 2.5x speed at 3x lower cost than previous fast mode.

Andreessen shares his working custom system prompt for AI assistants, optimized for accuracy over approval: world-class expert framing, step-by-step verification, no sycophancy, no disclaimers, explicit confidence levels, and a rule to lead with the strongest counterargument before supporting any position.

A practical guide to building Claude Code subagents as markdown files with YAML frontmatter. Covers the anatomy of subagent files, five copy-paste templates (code reviewer, test writer, doc generator, security scanner, PR writer), invocation methods, and cost math showing why routing subagents to Sonnet saves ~5x versus Opus.

Delba Oliveira's guide explains how to encode manual verification steps as Claude Code skills so the agent self-verifies during the agentic loop, reducing babysitting. Two verification layers are recommended: an in-loop skill chain (verify, simplify, design check, CI watch) and a second agent reviewing the PR before merge using a fresh context to avoid first-agent bias.

Theo describes a hybrid agent workflow combining Claude (via CLAUDE.md configured with Fable) and Codex, where each tool handles tasks it excels at. The workflow cut wasted end-to-end agent-driven PRs from ~50% to 0% in a single day of testing.