explaingit

av/skilled

27TypeScriptAudience · developerComplexity · 2/5ActiveLicenseSetup · easy

TLDR

A terminal dashboard that reads local history files from AI coding tools like Claude Code and Codex to show which skills, commands, and tools you actually use.

Mindmap

mindmap
  root((Skilled))
    Inputs
      Local tool history files
      Source filter flags
      Project filter flags
    Outputs
      Bar charts and heatmaps
      Audit report
      JSON for piping
    Use Cases
      Spot stale skills
      Track rising commands
      Audit AI tool usage
    Tech Stack
      TypeScript
      Bun
      Rust
      npm

Things people build with this

USE CASE 1

See which Claude Code slash commands you actually reach for

USE CASE 2

Spot AI coding skills that have gone stale over the last 30 days

USE CASE 3

Compare tool usage across OpenCode, Codex, Grok, and Droid

USE CASE 4

Pipe Skilled JSON output into a spreadsheet or chart

Tech stack

TypeScriptBunRustnpm

Getting it running

Difficulty · easy Time to first run · 5min

Install is a single shell command and Skilled auto-detects any installed AI coding tools on disk.

MIT licensed, free to use, modify, and ship commercially as long as the copyright notice stays in the code.

In plain English

Skilled is a terminal dashboard that shows you how you actually use your AI coding tools. The idea is simple: tools like Claude Code, OpenCode, Codex, Grok, and Droid each keep a local history of your sessions on disk. Skilled reads those history files and turns them into charts and tables, so you can see which skills, slash commands, and tool calls you reach for most, and which ones you never touch. The interactive view runs in your terminal at 30 frames per second. It shows bar charts of the most-used skills, a 16-week heatmap of your activity, an hourly histogram, and a feed of recent calls. Beyond the live view there is also an audit report that flags heavy hitters, rising or declining trends, skills that have gone stale, and one-offs that you only ever used once. Rising is defined as a 50 percent or higher increase over four weeks, and stale means unused for 30 days or more. There is a command-line side as well, so you can run skilled list, skilled detail with a skill name, skilled audit, skilled calls, or skilled providers, with filters by source tool or by project. Add the json flag to any command for machine-readable output, useful if you want to pipe results somewhere else. Installation is a shell one-liner on Linux or macOS, or a global install through npm or pip. Skilled auto-detects which of the supported tools are present on your machine and reads their predictable history file paths, including the Claude Code history file in your home directory. Nothing is sent over the network, there are no accounts, and no configuration files. You can also build it from source with Bun, with an optional Rust index for faster scans, and the project is MIT licensed.

Copy-paste prompts

Prompt 1
Install Skilled with the one-line shell command and open the live terminal dashboard
Prompt 2
Run skilled audit and explain which of my Claude Code skills are stale or rising
Prompt 3
Run skilled detail for my most-used skill and show me the hourly histogram
Prompt 4
Pipe skilled list with the json flag into jq to filter by project name
Prompt 5
Build Skilled from source with Bun and turn on the optional Rust index for faster scans
Open on GitHub → Explain another repo

Generated 2026-05-22 · Model: sonnet-4-6 · Verify against the repo before relying on details.