Skip to main content
plugin

claudestrophobic — Claude Code plugin: what it does, how to install

claudestrophobic is a Claude Code plugin that lists, inspects, renames, and deletes conversations and project data from the terminal. It clears orphaned sessions without waiting for the 30-day purge.

MIT ★ 10 updated 2026-09-23

Install into Claude Code · Plugin marketplace All methods and platforms ↓
/plugin marketplace add anthropics/claude-plugins-community
/plugin install claudestrophobic@claude-community
Open on GitHub

What claudestrophobic does

Claude Code lets you resume past chats, but leaves no native way to delete individual conversations. Everything sits under ~/.claude/projects/ as raw UUID filenames, and the built-in garbage collection only discards entries older than 30 days. The claudestrophobic plugin adds /sessions and /projects slash commands directly into your chat. You can inspect planned changes with --preview or remove chats by UUID prefix or description.

The plugin is written in a few hundred lines of dependency-free Python. Because it runs as skills rather than an MCP server, it consumes zero context tokens until called. When deleting a project, it clears subagent transcripts, memory, and history rows before running claude project purge. Active sessions remain safe because the script checks lock files. Running inside Claude Code's sandbox will block writes under ~/.claude, so deletion commands will fail there. Supported platforms are macOS and Linux.

What it helps with

  • delete a specific Claude Code conversation by UUID prefix or description
  • clean up orphaned project folders, memory, and history entries
  • preview which transcript and log files will be deleted before confirming
  • remove completed projects and trigger claude project purge automatically

Example prompts for the agent

  • List all conversations in the current project using /sessions
  • Delete a past chat about authentication: /sessions delete the one about auth
  • Clean up all orphaned project records with /projects delete orphaned --confirm

How to install claudestrophobic

Plugin marketplace

/plugin marketplace add anthropics/claude-plugins-community
/plugin install claudestrophobic@claude-community

Step by step, with commands for each platform: Claude Code plugins: marketplaces, installing, and building your own

Questions about claudestrophobic

How do I install claudestrophobic in Claude Code?

First run /plugin marketplace add anthropics/claude-plugins-community, then install the plugin with /plugin install claudestrophobic@claude-community.

What does claudestrophobic do?

It lets you inspect project sessions, remove individual chats by UUID prefix or description, and delete abandoned project data into the system Trash.

Does claudestrophobic add token overhead to regular chats?

No, it ships as skills rather than an MCP server. Its definitions are not attached to every prompt, so it uses zero context tokens until you invoke it.

Source text

SKILL.md · MIT
<div align="center">

# claudestrophobic

*Because `064ddd26-9a3f-...` is not a name.*

<a href="https://github.com/urav06/claudestrophobic/stargazers"><img src="https://img.shields.io/github/stars/urav06/claudestrophobic?style=social" alt="Stars"/></a>

<img src="https://img.shields.io/badge/Claude_Code-Plugin-5A67D8" alt="Claude Code Plugin"/>
<img src="https://img.shields.io/badge/license-MIT-green" alt="MIT License"/>
<img src="https://img.shields.io/badge/context_overhead-zero-black" alt="Zero context overhead"/>

<br>

<img src="https://github.com/user-attachments/assets/9f6ad16c-1b13-4336-81d5-faea0eb3eedd" alt="demo" width="680"/>

</div>

---

Claude Code lets you resume and rename past sessions, but not delete one. **claudestrophobic** adds that: delete a session, and find and remove the projects you have finished, all from the terminal. It ships as skills instead of MCP tools, so it costs **zero tokens** of context until you call it.

## Why

On claude.ai you can delete a conversation from the sidebar. Claude Code keeps the same conversations as UUID-named files under `~/.claude/projects/`, and the only built-in removal is a sweep that deletes everything older than 30 days.

This is that sidebar, brought to your terminal.

## Install

```
/plugin marketplace add urav06/claudestrophobic
/plugin install claudestrophobic@claudestrophobic-marketplace
```

## Usage

**`/sessions`** works inside the project you're in:

```
/sessions                             # list this project's sessions
/sessions delete the one about auth   # delete by description; the model finds the match
/sessions delete 064ddd26             # or by UUID prefix
/sessions delete 064ddd26 --preview   # show what a delete would remove
/sessions browse                      # open the project's folder
```

**`/projects`** works across all of them, from the outside:

```
/projects                             # every project, with its state: orphaned, dormant, or in use
/projects delete old-project          # preview what a delete would remove
/projects delete orphaned --confirm   # remove every orphaned project: sessions, memory, history, folder
```

## How it works

- **Names** come from each session's own transcript, the same chain Claude Code's resume picker uses: your `/rename` title, then the auto-generated one, then the first prompt.
- **Deletion** finds a session's files by name, anywhere under `~/.claude`, then removes only the ones in folders Claude Code itself treats as disposable: the transcript, subagent data, file history, session environment, debug log, and its rows in `history.jsonl`. It leaves anything found elsewhere alone and prints a pre-filled link to report it.
- **Projects** are keyed by their real working directory, read from a transcript, or from `history.jsonl` and `~/.claude.json` once the transcripts are gone.
- **Orphaned** means the project's directory no longer exists. Claude Code keeps its memory, history rows, and config entry forever, and nothing native notices. **Dormant** means the directory exists but no sessions do: the 30-day sweep took the conversations and kept the memory on purpose, and that memory loads the next time you run Claude there.
- **Deleting a project** removes every session, its memory, its history rows, and the directory, all to the Trash. It then runs Claude Code's own `claude project purge` to clear the project's entry in `~/.claude.json`, so it covers everything purge does and keeps the files recoverable. It refuses the project you are in and any with a live session.
- **Sandboxed sessions are told, not failed.** Claude Code's sandbox blocks reads of `history.jsonl` and `~/.claude.json` and writes under `~/.claude`. When that happens the list says which paths and states it couldn't verify, and a delete reports what the OS refused instead of stopping halfway. A `--confirm` only ever removes the set its preview showed; if that set changed in between, it refuses and previews again.
- **Active sessions are never touched.** They are read from Claude Code's lock files and refused. History rewrites are atomic, so a session running alongside never sees a half-written file. Everything removed goes to the system Trash where one exists, with `rm` as the fallback. macOS and Linux.
- **Zero tokens.** Skills load only when you call them. An MCP server with these features would add its tool definitions to every prompt you send; this stays out of your context until you ask. The whole thing is a few hundred lines of dependency-free Python.

© urav06 · MIT · urav06/claudestrophobic

Similar plugins

All plugins for Agent setup: memory, context, orchestration →

Project sponsors

Companies that keep this analytics open