Skip to content

Worktrunk is a CLI for Git worktree management, designed for parallel AI agent workflows

Docs: Note Claude Code user interrupt limitation

Maximilian Roosf62e44eea982
diff --git a/.claude-plugin/CLAUDE.md b/.claude-plugin/CLAUDE.md
index 07e5307db..140e57dd1 100644
--- a/.claude-plugin/CLAUDE.md
+++ b/.claude-plugin/CLAUDE.md
@@ -30,3 +30,16 @@ This approach keeps all Claude Code components organized together in `.claude-pl
**Note**: The official Claude Code documentation states "All other directories (commands/, agents/, skills/, hooks/) must be at the plugin root" but using the `source` field to point to `./.claude-plugin` makes paths relative to that directory, allowing this organizational structure.
**Why this works**: The `source` field in `marketplace.json` changes the base directory for path resolution. When `source: "./"` (the default), skills paths are resolved from the plugin root. When `source: "./.claude-plugin"`, skills paths are resolved from `.claude-plugin/`, allowing the entire plugin to be self-contained in one directory.
+
+## Known Limitations
+
+### Status persists after user interrupt
+
+The hooks track Claude Code activity via git config (`worktrunk.status.{branch}`):
+- `UserPromptSubmit` β†’ πŸ€– (working)
+- `Notification` β†’ πŸ’¬ (waiting for input)
+- `SessionEnd` β†’ clears status
+
+**Problem**: If the user interrupts Claude Code (Escape/Ctrl+C), the πŸ€– status persists because there's no `UserInterrupt` hook. The `Stop` hook explicitly does not fire on user interrupt.
+
+**Tracking**: https://github.com/anthropics/claude-code/issues/9516