Repository navigation
Releases
Skills Editor v0.3.0
Download the build for your platform below. Installed copies update themselves from here.
- Windows —
Skills.Editor_0.3.0_x64-setup.exe(or the.msi) - macOS —
Skills.Editor_0.3.0_universal.dmg(Apple Silicon and Intel) - Linux —
Skills.Editor_0.3.0_amd64.AppImage, or the.deb/.rpm
New in v0.3.0
macOS and Linux builds. Previously Windows-only. Path handling is now platform-neutral, and CI compiles and runs the full test suite on Windows, macOS and Linux so it stays that way. All artifacts are signed with the same updater key, so installed copies keep updating themselves on every platform.
macOS builds are not Apple-signed: the first open needs right-click → Open, or
xattr -dr com.apple.quarantine "/Applications/Skills Editor.app"
Windows is still the platform this is developed and tested on. The macOS and Linux builds pass the same automated tests, but haven't been exercised by hand — bug reports welcome.
Fixes
- Hook test runs no longer kill unrelated processes on macOS and Linux. When a test run hit its timeout, the cleanup signalled an entire process group; if the hook's command wasn't the group leader, that reached other processes. (On a CI runner it killed the build agent, which is how it was found.) It now only kills the command it started and that command's own children.
Skills Editor v0.2.0
Download and run the -setup.exe installer below. Installed copies update themselves from here.
New in v0.2.0
Hooks tab — browse every hook Claude Code loads (user and project settings, managed policy, plugins, skill/subagent frontmatter), edit them with an event-aware form or raw JSON, and see why one isn't running. Saving rewrites only the hooks key and refuses to clobber a file that changed on disk. Disable/enable individual hooks (Claude Code has no such switch, so a disabled hook is parked in the app's own storage), open the scripts a hook runs, and test-run a command hook with editable, event-shaped sample input — it reports the exit code, output, and how Claude Code would read the result.
Config tab — every file Claude reads as instructions: CLAUDE.md, .claude/CLAUDE.md, CLAUDE.local.md, rules (including paths:-scoped ones), parent-folder and subfolder files, @imports, and other agents' files (AGENTS.md, GEMINI.md, Copilot, Cursor, Windsurf, Cline). Each shows when it loads, and flags exclusions, broken imports, and over-long files. Startup context lists what a new session in a project reads before your first message, in load order, with a rough token count.
Memory tab — the notes Claude writes for itself: per-project auto memory and subagent memory. Notes are listed by name, type, description and age, with filters. It checks the MEMORY.md index against the folder: notes missing from the index (one click to add them), links pointing at files that are gone, and an index past its 200-line / 25 KB load limit. Notes can be created, edited with a name/description/type panel, and deleted along with their index line.
Sync now snapshots hooks, hook scripts, instruction files, and memory alongside skills.
Fixes
- Enabling a skill no longer reorders the entries in your
settings.json. - Spawned
gitandclaudeprocesses no longer flash a console window. gitruns with no terminal prompt, so a push or pull needing credentials fails fast instead of hanging forever.- Project paths no longer display with mixed
/and\separators.
Skills Editor v0.1.0
Download and run the -setup.exe installer below. Installed copies update themselves from here.