Local-first Skill management for every AI agent.
Skill-kits scans the exact local directories your agents read, then shows every Skill as enabled, disabled, invalid, or read-only across Codex, Claude Code, Gemini CLI, and custom agents.
skill-kits status$ skill-kits scan --project .
Agent reads it there.
Skill-kits manages it there.
Why it exists
Stop guessing which Skills your agents can see.
Modern AI coding setups span multiple agents, global folders, project folders, and plugin caches. Skill-kits gives that local file-system reality one precise control surface.
One file, whole tool
CLI and native GUI ship in the same skill-kits executable. Copy one binary, run it anywhere, remove it cleanly.
No runtime chain
No Electron, Node.js, Python, WebView, browser shell, or background daemon. Just a Rust binary reading local files.
Multi-agent visibility
See Codex, Claude Code, Gemini CLI, and custom Agent project spaces without merging distinct Skill copies into one vague record.
File-system truth
Enabled means SKILL.md. Disabled means SKILL.md.disabled. Invalid states are surfaced before they become workflow confusion.
Project-level inspection
Open Recent Projects, inspect project Skills, and enable or disable the selected instance without touching another agent's copy.
Codex plugin clarity
Codex plugins are shown separately from native Skills, with plugin status and read-only runtime capabilities kept explicit.
CLI and native GUI
A compact workbench for local Agent Spaces.
Use the CLI for repeatable checks and JSON output, or open the native GUI for a dense table-first view of Skills, plugins, agents, projects, paths, and status.
Enabled
The selected instance contains SKILL.md and is readable by that agent.
Disabled
The selected instance contains SKILL.md.disabled and can be safely toggled back.
Read-only
Plugin caches and vendor sources are visible without being rewritten in place.
Details
Built around the files agents actually read.
Skill-kits does not invent a cloud runtime for Skills. It keeps the real local directory model visible, inspectable, and reversible.
Bring order to your local AI Agent Skills.
Start with the CLI, open the native workbench when you need the full table, and keep every Skill operation tied to the path your agent actually reads.
$ cargo install --path .
skill-kits --version
skill-kits status
skill-kits --gui