Aedelon/claude-code-blueprint — reverse-engineered prompt

Build me a reusable Claude Code blueprint that I can copy into my own projects so Claude behaves more like a careful coding assistant instead of just guessing.

I want a clean repo with a main Claude instruction file, project rules for Python and TypeScript, a few useful skills, a few specialist agents, safety hooks, custom slash commands, and a settings template I can review before using. The goal is to make Claude research first, explain confidence, plan before editing, remember useful context between sessions, and avoid dangerous actions.

Please include hooks that can catch secrets, warn about risky shell or git commands, format code after edits, log failures, and summarize work at the end of a session. Add commands like agent, docs, and review. Include examples for skills such as anti hallucination, debugging, security audit, research, commits, and Python workflow.

Make the README clear enough that a normal developer can install it into their local Claude Code config safely. Look up current Claude Code docs online if needed.