Follow @plannotator on X for updates
Missed recent releases?
| Release | Highlights |
|---|---|
| v0.19.16 | Code navigation with peek view (Cmd/Ctrl+click tokens in diffs) |
| v0.19.15 | Commit-based diff base, jj evolution diffs, GitLab reliability fixes, OpenCode command intercept fix |
| v0.19.14 | Visual explainer skill update, PFM code-file hover previews, Graphviz, diff tab size and line bg intensity, hooks settings tab |
| v0.19.11 | Jujutsu (jj) VCS backend, slimmer hunk separators, collapse viewed files, multi-line gutter selection fix |
| v0.19.9 | OpenCode user-managed workflow, Pi model switch fix, Codex skill install, shimmer removal |
| v0.19.8 | 49 themes with syntax highlighting, keyboard shortcut registry, smart code-file path validation, remote URL notifications |
| v0.19.7 | Codex Stop-hook plan review, Codex skills, sidebar auto-close, file tree context menu |
| v0.19.6 | Non-blocking Pi browser sessions, agent picker dropdown for OpenCode, annotate-last file resolution fix |
| v0.19.5 | All-files diff view, clickable code file paths, server-side hide whitespace, non-ASCII path support |
| v0.19.4 | All-files diff type, code file viewer, hide whitespace, quick-settings popover |
What's New in v0.19.17
v0.19.17 reworks the goal setup skill from the ground up and adds a --version flag to the CLI.
Reworked Goal Setup Skill
The plannotator-setup-goal skill has been substantially rewritten. The previous version ran a rigid 5-document pipeline with sequential Plannotator review gates on a brief, plan, verification checklist, blockers list, and goal prompt. Each gate required a round trip through the UI before the agent could proceed, making the process slow and interruptive.
The new version replaces all of that with an interview-driven flow. The agent rearticulates your objective, then asks targeted questions to surface constraints, dependencies, and acceptance criteria. The answers are distilled into a fact sheet, which is the single Plannotator review gate. Once you approve the facts, the agent drafts the plan and produces the /goal prompt. Five gates become one; the Python scaffold script is gone; the agent writes files directly.
The result is a faster, more conversational process that produces better goal packages. The interview phase captures context that the old brief-writing step often missed, and reviewing a fact sheet is more concrete than reviewing a plan outline.
CLI Version Flag
plannotator --version (or -v) now prints the version and exits. The version is injected at compile time from package.json via Bun's --define flag in the release workflow, so compiled binaries report the exact release version. Uncompiled dev runs show plannotator dev.
Install / Update
macOS / Linux:
curl -fsSL https://plannotator.ai/install.sh | bashWindows:
irm https://plannotator.ai/install.ps1 | iexClaude Code Plugin: Run /plugin in Claude Code, find plannotator, and click "Update now".
OpenCode: Clear cache and restart:
rm -rf ~/.bun/install/cache/@plannotatorThen in opencode.json:
{
"plugin": ["@plannotator/opencode@latest"]
}Pi: Install or update the extension:
pi install npm:@plannotator/pi-extensionWhat's Changed
- feat: add --version / -v flag to CLI by @backnotprop in #725
- Simplify goal setup skill to interview-driven flow by @backnotprop in #727
Full Changelog: v0.19.16...v0.19.17