Decisions worth revisiting, not just the ones that worked.
A repeated Codex experiment comparing vanilla use, concise instructions, and AI coding frameworks, plus Claude Code research on success, runtime, tokens, and human intervention.
A practical permission policy for Claude Code and Codex, covering file edits, tests, package installs, network access, Git, secrets, and deployment.
A decision guide for Custom Agents, Skills, MCP servers, Hooks, and Subagents based on role, reusable workflow, external tools, and execution timing.
A comparison of studies on AGENTS.md efficiency, repository-context success rates, and configuration bloat, with rules for what belongs in AGENTS.md and CLAUDE.md.
A troubleshooting checklist for Claude Code MCP servers that show Connected while exposing no usable tools, covering /mcp, debug logs, Tool Search, paths, and permissions.