Claude Stuff

Things I know and find important to document about Claude (Code).

Some of it perhaps interesting to a wider audience, some of it clearly subject to my own preferences (see below).

Details

• Clojure as a primary language I am focusing on
• VSCode as my primary targeted IDE
• GitHub, Docker on the tooling side
• MacOS as my OS, Linux inside containers and remote

See also

• eighttrigrams/preferences

Topics

• Claude Code Configuration Files
• Claude and VSCode
• Claude Plugins
• Claude Marketplaces
• clojure-claude-and-mcp-knowledge.

Recipes

• How to Install Claude Code
• Working with MCP
• Docker Claude Container
• Writing a Hook Plugin That Rewrites a Tool Call

Issues

• Claude Shows Claude.ai Connectors in Claude Code

Eighttrigrams marketplace

See claude-marketplaces for background.

Plugins

clj-nrepl-eval

Evaluating Clojure code via nREPL.

• clj-nrepl-eval — nREPL SKILL copied wholesale from bhauman/clojure-mcp-light; install clj-nrepl-eval via bbin

architecture

Guidelines for structuring code.

• Architecture review
• Writing tests

log-tool-calls

A PreToolUse hook that logs every tool invocation to logs/hooks.log.

git-identity

A PreToolUse hook that stamps Claude <claude@eighttrigrams.net> onto any Bash call that commits, so agent commits are never signed as mine.

• Rewrites the command to export the four GIT_AUTHOR_*/GIT_COMMITTER_* variables on a line of their own, ahead of the original — a VAR=x cmd prefix would reach only the first command of a git add . && git commit
• Returns no permissionDecision, so it rewrites the command without granting it — "allow" would wave the whole line through, and "defer" hands the call back to the caller and discards the rewrite