What's wrong with this entry?
Anonymous. No account, no email.
What: Large tool outputs (>400KB) are now saved to disk instead of being truncated, allowing Claude to access the full data later.
How it works: When a tool produces output exceeding 400KB with ENABLE_TOOL_RESULT_SIZE_LIMIT enabled, the output is:
- Saved to
~/.claude/projects/<project>/tool-results/<tool>-<timestamp>/result.{txt|json} - A 2KB preview is shown inline
- Claude receives instructions on how to explore the full file
Example output:
<persisted-output>
Output too large (1.5 MB). Full output saved to: ~/.claude/projects/.../tool-results/Bash-1701234567890/result.txt
Preview (first 2.0 KB):
[preview content...]
...
You can explore this file using:
- Read tool to view portions of the file
- Grep tool to search for patterns
- Bash with jq to query JSON data
- Bash with head/tail for beginning/end
</persisted-output>
Details:
- Automatically detects JSON output and creates
.jsonfiles withjqusage examples - Tool result files are automatically granted read permission (no prompts needed)
- Falls back to truncation if file writing fails
- Evidence:
$_2()at line 403698,if5()at line 403760,nf5()at line 403783