Skilled

What is Skilled?

Skilled is a terminal user interface (TUI) dashboard that aggregates skill usage statistics across several AI coding tools, including Claude Code, OpenCode, Codex, Grok, and Droid. It reads local history and session files on the user's machine and turns them into a unified view of skill invocations, trends, and recent activity.

The project is designed for people who want to inspect how skills are being used across different AI coding environments without sending data to an external service. It provides both an interactive dashboard and command-line output for listing, auditing, and inspecting individual skills.

Key Features

• Cross-tool aggregation: Parses local traces from supported AI coding tools and normalizes them into a common format, making it easier to compare skill usage across providers.
• Interactive TUI dashboard: Shows bar charts, a 16-week activity heatmap, an hourly histogram, and a recent activity feed in a terminal interface.
• Skill audit views: Surfaces heavy hitters, rising and declining skills, stale skills, one-offs, and cross-project patterns to help users review usage at a glance.
• CLI commands with JSON output: Includes commands such as skilled list, skilled audit, skilled detail <skill>, skilled calls --source codex, and skilled providers, with --json available for machine-readable output.
• Filtering and sorting controls: Supports filtering by source or project, plus interactive sorting controls in the TUI for count, alphabetical order, and recency.
• Local-only operation: Reads history files locally and does not require accounts, network access, API keys, or telemetry.

How to Use Skilled

Install Skilled using the provided shell script, npm, or pip, then run skilled to open the dashboard. Once launched, you can browse the aggregated usage view, switch sort modes, open a skill's detail panel, or run CLI commands when you want focused reports or JSON output.

Use Cases

• Reviewing skill usage across tools: A developer who uses multiple AI coding assistants can see how often specific skills are triggered across different providers in one place.
• Finding inactive or overused skills: The audit view helps identify skills that are rising, declining, stale, or used only once, which is useful for housekeeping and workflow review.
• Inspecting one skill in detail: Users can open an individual skill to examine its usage pattern, source, and project breakdown instead of scanning a broad dashboard.
• Checking activity by project: Filtering by project makes it possible to understand which repositories or workspaces are driving particular skill invocations.
• Exporting data for scripts or analysis: The JSON output mode can feed other tools, scripts, or reporting workflows.

FAQ

Does Skilled send my data to a server? No. The source description says it reads local history files only and uses zero network and zero telemetry.

Which AI coding tools does it support? The page lists Claude Code, OpenCode, Codex, Grok, and Droid as supported or detected tools, with auto-detection based on whether their history files exist locally.

Can I use it without configuration? Yes. The documentation says Skilled auto-detects installed tools and shows them when their history files are present, so no manual configuration is needed in typical use.

Does it provide a command-line interface as well as a TUI? Yes. The project includes both an interactive terminal dashboard and CLI commands such as list, audit, detail, calls, and providers.

Can it work with large history files? The repository mentions an optional Rust index for faster re-scanning of large history files, which the TUI uses automatically when available.

Alternatives

• Tool-specific usage logs: Each AI coding tool may already keep its own local history, but those logs are limited to a single provider and do not unify usage across tools.
• Custom scripts over local JSONL files: A developer can write scripts to parse session and history files directly, but that requires maintaining the parsing and aggregation logic themselves.
• General terminal dashboards: Broader TUI dashboards can visualize activity, but they may not understand AI coding skill traces or the specific concepts Skilled extracts.
• Web-based analytics services: Cloud analytics products can offer dashboards and reporting, but they typically rely on uploading data, which differs from Skilled's local-only workflow.