Five Claude Code Mods worth trying
Small TypeScript functions that run inside Claude Code and can change your prompt before the model reads it. Mods are powerful and unsandboxed: they can access anything Claude Code can. Read the source before you install.
Anthropic just let developers rewrite Claude Code from the inside.
Mods are small TypeScript functions that run inside Claude Code and can change your prompt before the model ever reads it.
One hides your secrets first, so the model never sees your keys. Another won't let Claude write code until the tests exist.
But here's the catch: Mods aren't sandboxed. They can touch anything Claude Code can. Even Anthropic's own /diff pane is a Mod, so these are worth using carefully.
Official Mods site: https://claude.dev/mods
Five Mods worth trying Read each project's source before installing. Get the list on the website, or comment "MODS" and we'll send it to your DMs.
1. claude-code-redact https://github.com/karanb192/claude-code-redact
Hides API keys and personally identifiable information from the model prompt, replacing them with placeholders.
2. tdd-mod https://github.com/koslowskyj/tdd-mod
A judge model checks file writes and blocks them, with a reason, if they skip a failing test or implement more than the test needs. Marked experimental; built and tested on Claude Code 2.1.288. It blocks the write if the judge can't be reached.
3. agent-flow https://github.com/Charlie0113-T/claude-agent-flow
/flow opens a live tree of the session's sub-agents beside the transcript.
4. aside https://github.com/JayDoubleu/aside
/aside opens a read-only side chat in a pane. You ask about the session so far, and a copy of the transcript answers, so your main thread isn't touched.
5. cctop https://github.com/tomstagl/cctop
A btop-style live dashboard in a side pane. It shows context fill, tokens, cost, rate limits and per-tool latency.