| Name | Modified | Size | Downloads / Week |
|---|---|---|---|
| Parent folder | |||
| caprock_0.39.1_windows_amd64.zip | 2026-08-31 | 8.8 MB | |
| caprock_0.39.1_windows_arm64.zip | 2026-08-31 | 8.0 MB | |
| checksums.txt | 2026-08-31 | 600 Bytes | |
| caprock_0.39.1_darwin_amd64.tar.gz | 2026-08-31 | 8.7 MB | |
| caprock_0.39.1_darwin_arm64.tar.gz | 2026-08-31 | 8.2 MB | |
| caprock_0.39.1_linux_amd64.tar.gz | 2026-08-31 | 8.6 MB | |
| caprock_0.39.1_linux_arm64.tar.gz | 2026-08-31 | 8.0 MB | |
| README.md | 2026-08-31 | 1.1 kB | |
| v0.39.1 source code.tar.gz | 2026-08-31 | 5.6 MB | |
| v0.39.1 source code.zip | 2026-08-31 | 5.9 MB | |
| Totals: 10 Items | 61.7 MB | 0 | |
Fixed
- Shift+Enter typed
\ninto the prompt and sent the message. The newline keys sent two printable characters — a backslash and the letter n — so a message arrived asfirst line\nand the multi-line prompt was unusable. They now sendESC CR, which is Alt+Enter as a terminal encodes it and what Claude Code's own macOS instructions bind Option+Enter to.
Three earlier answers were wrong, and two of them shipped. CSI u needs
the kitty keyboard protocol negotiated, which we never do. 5c 6e came from
reading the binding /terminal-setup writes into iTerm2 — but Send Text
interprets that escape, so iTerm2 puts one byte on the wire rather than
two characters. A bare line feed looked correct against an empty prompt
and submits the moment anything is typed, which is the symptom that was
reported in the first place.
Testing on an empty prompt is what made two wrong answers look right. The sequence is now verified by sending candidate bytes to a running Claude Code with text already typed, and the tests assert the bytes rather than the reasoning.
