Blog & Skills

Notes, opinions, and copy-paste-ready skills.

context-mode: stop your coding agent from re-reading everything it already saw

An MCP server + hooks setup that keeps raw tool output out of your context window — 315 KB to 5.4 KB — and rebuilds session memory from SQLite instead of a context dump.

Aug 6, 2026 · 2 min read

Terminal dashboard showing real-time token attribution with quota bar, burn rate, and per-source breakdown by tool, agent, skill, and MCP server

skill · curated

claude-token-lens: see exactly what's burning your Claude quota, live

A live terminal dashboard that attributes every token to a specific tool, skill, agent, or MCP server — so you know what to fix before you hit the wall.

Aug 3, 2026 · 2 min read

skill

skill · curated

obra/superpowers: the coding-agent framework that makes your AI plan before it codes

Jesse Vincent's obra/superpowers enforces a seven-phase workflow — brainstorming through branch finishing — so your agent actually designs before it ships. 258k GitHub stars, MIT licensed, works across eight coding agents.

Jul 20, 2026 · 2 min read

dev-browser header image — browser automation for Claude Code

skill · curated

dev-browser: Give Claude a Playwright-backed browser without handing it your filesystem

A QuickJS-sandboxed Claude Code skill that wraps the full Playwright API — so your agent can click, fill, and screenshot without touching the rest of your system.

Jul 13, 2026 · 2 min read