qcode v1.2.1

qcode v1.2.1 adds guided skill planning and strengthens prompt-injection handling, while improving agent responsiveness and conversation exports.

Highlights

Added /skillplan, a guided skill-authoring mode that creates a reviewable draft through focused questions. Drafts can be refined before /skillplan create writes a validated SKILL.md to a supported location.
Added trust boundaries around tool output and imported context, with warnings for actionable injected instructions and exact per-call approval before subsequent data access, network requests, or mutations.
Kept agent tabs responsive during /compact by queuing compaction with prompts and showing progress and completion in the originating tab.
Improved pretty conversation exports with clearer colors and completed responses folded under their prompts.
Expanded agent, skill, trust, session, and TUI test coverage.

Download the archive matching your platform, extract qcode (qcode.exe on Windows), and place it somewhere on your PATH.

tar -xzf qcode_linux_amd64.tar.gz
./qcode --version

Use SHA256SUMS to verify downloaded archives:

sha256sum -c SHA256SUMS

Standalone platform binaries are also included for the qcode update command.