disk-cleanup

To recover from ENOSPC disk full errors when Bash is deadlocked, truncate selected large text files in-place using Read and Write operations to free space, then use safe find-and-remove patterns on build artifacts and tmp caches once Bash is available again.

Your root filesystem shows 100% usage and Claude session initialization fails because disk exhaustion prevents writing session-env files. Wedged session initialization and accumulated build or test artifacts can cause Bash tool invocations to fail before running any command.

The best way to reclaim disk space from build artifacts and tmp emitter caches is applying safe find-and-remove patterns to typical compiler and test directories, clearing large /tmp caches, and validating recovery with df -h /.

To truncate large text files when shell access is unavailable due to no space left on device, use Read to identify the files, then Write with empty content to truncate them in-place without relying on Bash execution.

Yes, you can diagnose deleted-but-still-open files after disk cleanup using lsof and /proc to identify processes still holding file handles, which can prevent reclaimed space from registering on the root filesystem.

Yes, this disk cleanup approach works for lyric-lang dev environments specifically, targeting the exact build artifacts, test directories, and tmp emitter caches generated by the dotnet build and test workflows in that environment.