ag-skill
ag-skill is a command-line tool for installing and managing AI engineering assets — skills, templates, and agent bundles — directly in your repository. Drop in battle-tested domain knowledge for your AI agent in seconds, keep it version-controlled alongside your code, and upgrade it as the catalog evolves.
What it does
| Command | Description |
|---|---|
ag-skill list | Browse all skills available in the catalog |
ag-skill install <skill> | Install one or more skills into your repo |
ag-skill add <bundle> | Scaffold a full agent bundle (e.g. core-agents) |
ag-skill remove <skill> | Remove installed skills |
ag-skill upgrade | Upgrade installed skills to the latest catalog version |
ag-skill doctor | Validate installed skills against the supported spec |
ag-skill observe | Record and review local skill/agent run telemetry |
ag-skill stats | Print aggregated token usage and success-rate statistics |
ag-skill dashboard | Generate a self-contained HTML telemetry dashboard |
Highlights
DDD skills
The catalog now includes a broader Domain-Driven Design suite that covers strategy, tactical modeling, and implementation orchestration:
ddd-strategic-design— classify subdomains, draw bounded contexts, and choose integration patterns before writing any code.ddd-aggregate-design— model aggregate boundaries and invariants with single-transaction consistency.ddd-application-services— design thin use-case handlers that orchestrate aggregates without leaking domain logic.ddd-repositories— apply repository boundaries and clean persistence abstractions.ddd-cqrs— separate write-side aggregate workflows from read-side projection/query models.ddd-domain-services— place cross-aggregate business logic in stateless domain services.ddd-sagas— coordinate long-running, compensating workflows across bounded contexts.
Together these skills produce practical artifacts (context-map.md, ADRs, handler/repository patterns) that carry cleanly into implementation.
Core Agents bundle
The core-agents bundle scaffolds a four-role AI agent team — architect, coder, UI developer, and tester — tuned for the .NET / ASP.NET Core + React stack. Install it once and your agents share consistent conventions across every project in the repository.
ag-skill add core-agents
See the Core Agents bundle → for the full role breakdown and workflow.
Local telemetry — observe, stats, dashboard
ag-skill tracks skill and agent runs entirely on disk — no external service, no account required. After your agent finishes, record the run:
# Auto-read the latest VS Code Copilot Chat transcript
ag-skill observe from-copilot --skill ddd-aggregate-design --agent ag-architect --model gpt-4o --duration-ms 12400
# Auto-read the latest Claude Code session (exact token counts)
ag-skill observe from-claude --skill ddd-aggregate-design --duration-ms 18500
Then review with ag-skill stats for terminal tables or ag-skill dashboard for an HTML report. Everything lives in .ag-skill/telemetry/runs.jsonl inside your repo.
See the Local Telemetry guide → for the full workflow.
Supported agent targets
Install skills for the agent tool of your choice:
copilot— GitHub Copilotclaude— Claudecursor— Cursorall— all targets at once
Requirements
License
ag-skill is released under the Apache License 2.0.